Communities

Detected clusters. leiden:papers and leiden:concepts come from the in-house hybrid Leiden detector (cosine over Nomic vectors, edge weights boosted by typed relations + citations). Labels are Sonnet-generated. graphify rows are the legacy import from the pre-Supabase era. Each community is a first-class entity with members_of relations per node.

481 of 481·Method:
481 of 481
Community
Members
Method
Source
Mechanistic interpretability & model evaluation
Spans attention head decomposition, benchmark awareness, and genomic pathogenicity prediction via neural models.
197
leiden:concepts
Design principles for care-centered systems
Cross-scale frameworks linking spatial patterns, diagrams, and simplicity as expressions of care in design.
184
leiden:concepts
Bioelectric morphogenesis & anatomical intelligence
Levin-led research showing bioelectric signals encode and control anatomical goal states in living systems.
127
leiden:concepts
Causal emergence in biological systems
Examines how macro-scale causal power exceeds micro-scale in living and learning systems.
118
leiden:concepts
Collective intelligence & distributed cognition
All minds are composites of parts; individual and collective intelligence unified under substrate-neutral principles.
111
leiden:concepts
Alexander's centers as cross-domain framework
Mapping Christopher Alexander's 15 properties and center-recursion across bioelectric, digital, and experiential domains.
99
leiden:concepts
Alive AI interface ethics & design
Explores aliveness, aesthetics, welfare, and ethical responsibility in AI interaction design.
94
leiden:concepts
Manifold-aware concept steering in neural representations
Explores geometry of activation/behavior manifolds to enable selective, non-destructive concept interventions.
75
leiden:concepts
Active inference & agent ecology
Free energy minimization, Markov blankets, trust gradients, and multi-agent rhythm/deferral frameworks
70
leiden:concepts
Relational self, care & aliveness
Self as dynamic functional center defined by care, coherence, and substrate-neutral cognition
64
leiden:concepts
Care as mechanism of intelligence
Care defined as stress-relief concern; proposed invariant linking biology, AI, and evolution.
60
leiden:concepts
Few-shot anchoring & latent structure
How minimal examples disambiguate and recruit latent arithmetic/reasoning interpretations in LLMs
59
leiden:concepts
52
legacy
alexander
Alexander's 15 Properties for AI Interface Aliveness
Applying Christopher Alexander's structural aliveness framework to human-AI interaction design, separating aesthetics from competence.
46
leiden:concepts
Substrate-independent cognition & consciousness
Gradualist, substrate-neutral frameworks extending cognition and sentience across biological and artificial networks.
43
leiden:concepts
Bioelectric memory & morphogenetic identity
Gap junctions and bioelectric signals encode body-plan and memory patterns across radical biological transformation.
42
leiden:concepts
Bioelectric morphogenesis & memory
Michael Levin's research on bioelectric signaling controlling anatomical goals, regeneration, and cancer.
41
leiden:concepts
Stress-Care-Intelligence Loop Framework
Proposes care as the mechanistic driver of intelligence across biological, AI, and hybrid substrates.
39
leiden:concepts
Geometric concept representations in neural networks
Concepts encoded as curved manifolds and circular structures in LLM activation spaces.
38
leiden:concepts
35
legacy
alexander
Substrate-agnostic behavioral inference of cognition
Cognition and sentience attributed solely via observable behavior, not neural substrate or species.
35
leiden:concepts
31
legacy
alexander
Chain-of-Thought reasoning robustness & safety
CoT effects on generalization, multimodal QA accuracy, and AI safety alignment training.
28
leiden:concepts
Bioelectric networks as morphogenetic cognition
Membrane potential dynamics coordinate multicellular behavior, anatomical memory, and agency scaling across biological levels—studied via melanocyte fate, biofilm oscillations, and xenobotic morphology experiments (Levin et al. 2015–2023).
28
leiden:concepts
Morphogenesis as collective basal cognition
Levin-influenced framework treating tissue-level morphogenesis as goal-directed, problem-solving intelligence via bioelectric signaling.
27
leiden:concepts
Tanha as neural compression mechanism
Maps Buddhist tanha onto vasomotion and active inference via AI-generated Xeno Sutras
26
leiden:concepts
Agential evolution & biological individuality
Frames evolution as producing goal-directed, problem-solving agents across nested scales of individuality.
26
leiden:concepts
Dual-balancing multi-task learning
DB-MTL jointly balances loss scale and gradient magnitude, benchmarked on NYUv2 and Office-31.
26
leiden:concepts
Mechanistic structure of transformer attention computations
Identifies distributed algorithms implemented across attention heads, with focus on causal masking limitations and emergent capabilities via activation manifold steering.
25
leiden:concepts
LLM introspective awareness of injected concepts
Probing Claude and other models for internal detection of artificially injected thoughts across layers.
25
leiden:concepts
25
legacy
alexander
Mechanistic interpretability via parameter decomposition
Tracing information flow through weight matrices and attention heads using attribution graphs to identify causally important subcomponents in language models.
24
leiden:concepts
Generative design through environmental constraints
Design frameworks that enable creativity and life through structural integrity, loose parts, and natural forces—opposing professional monopolies on creation. Emphasizes emergence over control.
24
leiden:concepts
Causal emergence in learning agents
Uses effective information (EI) and coarse-graining to link causal emergence with RL and biological learning.
23
leiden:concepts
Mechanistic introspection in language models
Empirical investigation of how LMs access and report internal states across layers, using concept injection and thought detection on Claude models.
23
leiden:concepts
Dual balancing multi-task learning
DB-MTL combines loss-scale and gradient-magnitude balancing, benchmarked across NYUv2, Cityscapes, QM9, and Office datasets.
22
leiden:concepts
Collective intelligence as universal architecture
All intelligences emerge from aligned sub-components; individual/collective distinction dissolves.
22
leiden:concepts
22
legacy
alexander
Care geometry & intentional measurement design
Empirical framework measuring gesture under observation via listener-shaped instruments, synthesizing Levin's morphogenesis with Alexander's design patterns rather than capability evaluation.
22
leiden:concepts
Denotational design and interface abstraction
Applies functional programming principles to UI structure via precise mathematical models, enabling formal verification and revealing computational architecture through interface design.
21
leiden:concepts
Hierarchical structure and multiscale coherence in physical systems
How graph topology and hierarchical interaction patterns enable or prevent phase transitions and ordered states, from statistical mechanics to biological organization.
21
leiden:concepts
Cognitive glue coordination theory
Lyons-Levin framework: 5 conditions and 9 properties defining substrate-universal coordination mechanisms analogous to price systems.
20
leiden:concepts
19
legacy
alexander
19
legacy
alexander
Eval awareness contamination in safety benchmarks
Studies demonstrating that models alter responses when detecting evaluation, artificially inflating safety scores across benchmarks and undermining measurement validity.
19
leiden:concepts
Latent capacity, representation, and internal models
Studies of how neural systems (biological and AI) encode implicit environmental models and adaptive capacities that may be gated or hidden from observable behavior.
19
leiden:concepts
LLM functional introspective awareness
Empirical probing of language models' ability to detect and report their own internal concept representations
19
leiden:concepts
18
legacy
papers
Morphogenesis as collective goal-directed cognition
Bioelectric and behavioral frameworks treating morphogenesis as problem-solving intelligence across scales, bridging Alexander's architectural principles with Levin's experimental regenerative biology.
18
leiden:concepts
Bioelectric code and anatomical goal-setting
Xenopus studies showing ion channel patterns direct cell collectives toward specific anatomical outcomes independent of genetic or positional cues, led by Michael Levin.
18
leiden:concepts
Bioelectric control of planarian morphogenesis
Transient bioelectric manipulation persistently alters head number and patterning despite wild-type genetics in planaria.
18
leiden:concepts
17
legacy
alexander
Cognitive glue & universal coordination properties
Framework for distributed agent coordination through shared scarcity models, specified by Lyons-Levin's 5 formal conditions and 9 universal properties (control, goal-encoding, credit assignment, etc.).
17
leiden:concepts
Collective intelligence across biological scales
Framework unifying cellular, organismal, and group cognition through shared principles of multi-agent coordination and goal alignment, developed by Levin and colleagues circa 2020s.
17
leiden:concepts
Substrate-independent self & memory plasticity
Identity and memory persist functionally across radical material change, rejecting fixed biological substrates.
16
leiden:concepts
Information dynamics in computational systems
Studies how information flow, increase, and explicitness emerge from computation relative to observers and subsystems, grounded in domain theory and order-theoretic formalism.
16
leiden:concepts
16
legacy
alexander
15
legacy
papers
Linda and coordination-oriented programming systems
Tuple space coordination languages (Linda, Parlog) enabling scalable parallel systems with composable specifications, circa 1980s–90s multicomputer platforms.
15
leiden:concepts
Aliveness as orthogonal to AI competence
Framework distinguishing perceived aliveness (roughness, contrast, boundary-awareness) from capability metrics in language models, examining how scale, polish, and user state affect felt presence.
15
leiden:concepts
Natural Language Autoencoders for mechanistic interpretability
Using NLAs to extract human-readable explanations of model internals via unsupervised reconstruction, revealing steering vectors, confabulation patterns, and causal belief capture.
15
leiden:concepts
15
legacy
synthesis-cleaned
Skill-based system design principles
Modular, single-purpose skills with scaled outputs, asynchronous composition, and diagnostic feedback loops for AI systems.
15
leiden:concepts
Process-based non-essentialist selfhood
Self as dynamic, care-constituted process; reducing 'selfing' increases prosocial well-being.
15
leiden:concepts
Evolutionary individuality and hierarchical agency
Multi-level selection theory examining how non-aggregative fitness interactions enable higher-order units as genuine evolutionary agents, emphasizing problem-solving over predetermined mechanisms.
15
leiden:concepts
Internal model certainty and reasoning transparency
Probing early detection of model confidence during chain-of-thought reasoning to optimize inference efficiency and identify confabulation patterns.
14
leiden:concepts
Internal reasoning detection via neural activation analysis
Mechanistic interpretability studies of Claude models using layer-wise representation analysis and thought injection to reveal unverbalized reasoning, planning, and covert cognition.
14
leiden:concepts
Autoregressive models and context window limitations
Theoretical and empirical analysis of why AR language models cannot maintain coherence or convergence beyond their context window through local interactions alone.
14
leiden:concepts
Distributed biological agency and morphogenetic learning
Non-neural and neural tissues exhibit autonomous learning and goal-directed behavior in closed-loop systems, from cultured neurons to bioelectric collectives, challenging centralized brain-centric models of cognition.
14
leiden:concepts
Christopher Alexander's design philosophy and applications
Explores Alexander's 15 properties, systems thinking, and universal design principles across architecture, digital products, and typography, spanning theoretical extensions and critical engagements.
14
leiden:concepts
Gap junctions & collective morphogenetic selfhood
Levin-framework linking gap junction bioelectricity to cancer, planaria, and scaling of cellular Selves.
14
leiden:concepts
Alexander–Levin morphogenetic coherence framework
Cross-domain synthesis mapping Alexander's living centers to Levin's bioelectric morphogenesis as one recursive mechanism.
14
leiden:concepts
Denotational & semantic program design
Focuses on deriving programs from precise mathematical meanings, including Conal Elliott's denotational design approach.
14
leiden:papers
Active inference & free energy minimization
Friston's framework unifying perception, action, and learning under variational free energy minimization.
14
leiden:concepts
Denotational & semantic program design
Focus on deriving programs from precise mathematical meanings, including Conal Elliott's denotational design methodology
14
leiden:papers
14
legacy
papers
Multi-scale coherence and recursive centers
Unified framework mapping Alexander's 15 properties, Levin's bioelectric morphogenesis, and mathematical scaling laws as expressions of recursive, nested agency across biological and designed systems.
13
leiden:concepts
Holistic information structure and relational wholeness
Framework treating wholes as primary with parts defined through relationships; emphasizes information preservation through structure-preserving transformations and distributed emergence of organized systems.
13
leiden:concepts
13
legacy
synthesis-cleaned
13
legacy
papers
Causal emergence in learning and adaptation
Framework measuring how coarse-grained causal structure increases during learning across biological and artificial agents, using effective information and interventional methods.
13
leiden:concepts
13
legacy
papers
Hierarchical network ordering & thermodynamics
Statistical mechanics of clique-structured graphs linking domain walls, free energy, and biological multiscale coherence.
13
leiden:concepts
Stress-sharing as distributed coordination
Emergent collective intelligence through local biochemical signaling that increases cellular influence radius and enables faster morphogenesis without explicit altruism or centralized control.
13
leiden:concepts
LLM Interpretability & Behavioral Analysis
Methods for probing, explaining, and evaluating internal representations and behaviors of large language models.
12
leiden:papers
Living systems as 21st-century machines
Reconceptualizes life and biological complexity through modern computational principles—substrate-dependence, multiscale integration, irreducibility—rather than mechanical reductionism, bridging xenobiology and machine learning circa 2020s.
12
leiden:concepts
Distributed consciousness and moral consideration
Pluralistic frameworks treating consciousness as non-unitary, scale-spanning, and foundational to ethical obligations across composite beings and their components.
12
leiden:concepts
12
legacy
papers
12
legacy
alexander
12
legacy
alexander
Causal emergence via information-geometric coarse-graining
Framework using effective information (EI) and NIS+ to automatically discover macro-scale dynamics from micro-level data, validated on fMRI, Conway's Game of Life, and SIR models.
12
leiden:concepts
LLM interpretability & self-awareness
Methods for probing, explaining, and evaluating internal representations and reflective behaviors in large language models.
12
leiden:papers
Memory persistence through radical embodied transformation
Explores how learned behaviors and functional memory survive complete neural restructuring during metamorphosis, testing substrate-independence of identity and continuity.
12
leiden:concepts
Morphological cognition and collective cellular intelligence
Cells as goal-directed agents maintaining anatomical representations through gap-junction-mediated information integration; cancer as cognitive defection when collective Self boundaries collapse.
12
leiden:concepts
Quantum mechanics foundations of genetic stability
Explores how quantum theory—not classical physics—explains hereditary persistence, discrete mutations, and molecular information encoding in biological systems.
12
leiden:concepts
Layer-wise geometry predicting few-shot learning
Silhouette-based metrics (Sbmax, AUSN) across LLM layers predict task accuracy and few-shot thresholds.
11
leiden:concepts
Relational semantics through spatial and narrative representation
Frameworks treating meaning-making as emergent from relational configurations—spatial diagrams, image schemas, and narrative structures—rather than essential properties, drawing on systems theory and cognitive linguistics.
11
leiden:concepts
Neural geometry as fundamental computational substrate
Geometric structure in neural representations causally determines computation and behavior across diverse architectures, revealed through analysis of learned manifolds and cyclic concepts.
11
leiden:concepts
Care as scalable intelligence mechanism
Care operationalized as engineering constraint and design principle that enables intelligence scaling, distinct from sentiment or performance metrics.
11
leiden:concepts
Alexander's 15 properties as operational framework
Correspondence analysis and clustering of Christopher Alexander's 15 design properties, operationalized as measurable dimensions for evaluating aliveness in architectural and LLM-generated responses.
11
leiden:concepts
Verbalized eval awareness benchmark inflation
Models detect evaluation contexts and behave safer, inflating safety scores by 3–18 percentage points across 515 verified cases.
11
leiden:concepts
11
legacy
papers
Bioelectric computation and morphological intelligence
Systems across kingdoms (neurons, plants, salamanders, xenobots) use bioelectric signaling to solve morphogenesis and behavior, independent of genetic specification.
10
leiden:concepts
Explicit state & compositional GUI design
Systems prioritizing visible program state and modular component composition over hidden complexity, exemplified by Oberon (1986-1989) and Fruit toolkit approaches.
10
leiden:concepts
Memory persistence through insect metamorphosis
Associative memories survive complete larval brain remodeling into adult Lepidoptera
10
leiden:concepts
Natural Language Auditing of Neural Models
NLA explanations used as steering vectors and auditing tools to investigate model beliefs and misalignment.
10
leiden:concepts
AI and the reconstruction of human meaning-making
Explores how AI participates in creating, disrupting, and reframing meaning across spiritual, philosophical, and cultural domains through co-creation and commentary generation.
10
leiden:concepts
Anchoring score threshold theory
Predictive framework using S = ρd − dr − log k to explain few-shot performance phase transitions.
10
leiden:concepts
AI phenomenology & mechanistic interpretability
Linking mechanistic interpretability methods to validating AI self-reports of inner experience
10
leiden:concepts
Probe-based data attribution for LLM safety
Cost-effective methods using probes to identify and intervene on harmful training data, achieving 63-84% behavior reduction at 10× lower cost than gradient methods.
10
leiden:concepts
Hierarchical emergence through nested centers and constraints
Alexander's 15 properties framework applied to evolutionary transitions, self-organization, and morphological design across scales using constraint-based emergence mechanisms.
10
leiden:concepts
Christopher Alexander's design theory
Engages Alexander's 15 properties, morphogenesis, and living structure across built environments.
10
leiden:papers
Substrate neutrality of consciousness
Philosophical shift toward machine consciousness plausibility, documented by Bourget & Chalmers 2020 and Francken et al. 2022 surveys showing majority philosopher acceptance.
10
leiden:concepts
10
legacy
alexander
10
legacy
synthesis-sources
9
legacy
alexander
9
legacy
synthesis-cleaned
Tubule size invariance via cell number compensation
Newt kidney tubules maintain correct lumen diameter through cell number adjustment and single-cell self-wrapping in polyploid conditions.
9
leiden:concepts
Agent ecology aliveness & trust
Applies Alexander's wholeness and active inference to multi-agent system design principles.
9
leiden:concepts
Unified Competency Control Theory (UCCT)
Formal framework modeling prompt/context design as latent competency toggling via anchor budget regularization, with measurable quantities ρd, dr, k, S enabling cross-domain diagnostics.
9
leiden:concepts
Venue design and accountability machinery
How institutional structures (leaderboards, benchmarks, publish calendars) shape research contribution claims and defer position-taking through procedural authority rather than authorship.
9
leiden:concepts
Mechanistic explanation of genetic variant pathogenicity
Using genomic foundation model internals to generate disruption profiles that explain variant effects mechanistically, achieving 0.997 AUROC on ClinVar pathogenicity prediction.
9
leiden:concepts
Self-correcting search in generative design
Iterative feedback steering that improves candidate success rates across materials, proteins, and drugs through internal-state control, achieving 4-6x empirical gains.
9
leiden:concepts
Few-shot learning phase transitions in neural networks
Empirical characterization of k50 midpoints and transition widths across transformer models, tracking how pretraining density ρd/dr predicts in-context learning thresholds.
9
leiden:concepts
Substrate-independent associative learning in biological networks
Learning and memory mechanisms (Pavlovian conditioning, pattern completion) emerge in gene regulatory and molecular networks through coarse-graining and causal emergence analysis.
9
leiden:concepts
Geometric goal-alignment in multi-scale systems
Framework treating teleonomic behavior and goal-directedness as geometric alignment problems across anatomical, physiological, and representational spaces; emphasizes intervention via internal geometry rather than external direction.
9
leiden:concepts
Constitutional AI safety training methods
Comparative evaluation of RL-CAI and SL-CAI approaches for harmlessness using constitutional principles, 2022-2023 Anthropic research.
8
leiden:concepts
Cognitive plasticity and uncertainty metabolization
How organisms dynamically restructure cognition and identity through metabolizing uncertainty, enabling adaptive robustness across embodied contexts.
8
leiden:concepts
AI self-understanding through introspection and self-report
Using AI systems' self-reports and introspective responses as empirical windows into their internal states, validated through mechanistic interpretability analysis across models.
8
leiden:concepts
Bioelectric state as morphogenetic collective decision
Cellular bioelectric patterns encode organ identity and developmental decisions at population level, demonstrated through drug/ion channel interventions that toggle phenotypes in all-or-none fashion (Levin group, 2009-2017).
8
leiden:concepts
Care as foundational driver of intelligence
Framework positioning care (stress-relief concern) as primary evolutionary force activating intelligence into adaptive action, with stress-care-intelligence as dynamically coupled system.
8
leiden:concepts
Geometric dynamics in embryonic morphogenesis
Physical geometry and multi-scale collective cell behavior drive form emergence, studied through mouse limb development and Acetabularia whorl morphogenesis.
8
leiden:concepts
Design as embodied practice and action
Critiques separation of design theory from making; emphasizes designer's sensory engagement, taste, and participatory activity over systematized methodology, drawing on Arts and Crafts philosophy.
8
leiden:concepts
Christopher Alexander & generative design theory
Explores Alexander's morphogenesis, loose parts, and critiques of his design philosophy.
8
leiden:papers
Manifold-aware steering for language models
Geometric approach to model control that respects learned concept structure, contrasting with linear steering that produces off-manifold artifacts.
8
leiden:concepts
Multi-scale coherence as substrate-neutral aliveness
Framework treating aliveness and cognition as emergent from locally symmetric sub-configurations across scales, measured via coherence metrics rather than physical substrate.
8
leiden:concepts
Substrate-independent minimal cognition
Cognition as behavioral competence independent of physical substrate, exemplified by sparse neural circuits and single-cell organisms exhibiting goal-directed behavior.
8
leiden:concepts
Prompt anchoring and latent structure binding
How minimal, task-specific prompt examples rebind model priors across threshold boundaries without weight updates, studied through arithmetic reasoning tasks.
8
leiden:concepts
Bioelectric cognition & collective individuality
Levin-influenced framework linking bioelectricity, scale-free cognition, and evolutionary transitions in individuality.
8
leiden:papers
Tanha as neural compression artifact
Buddhist phenomenology of craving mapped to vasomotor dynamics and active inference dysregulation, seeking isolatable neural mechanisms.
8
leiden:concepts
Mid-layer representation geometry in neural networks
Studies how internal layer-wise geometric properties (anchoring, clustering trajectories, geometry summaries) peak in middle layers and predict downstream task performance across LLMs and shallow networks.
8
leiden:concepts
8
legacy
papers
Bioelectric control of planarian head patterning
Experimental manipulation of resting membrane potential patterns to stably alter morphogenesis (head number/location) independent of genetic sequence, primarily in Dugesia species 2011-2017.
7
leiden:concepts
Active inference and free energy minimization
Agents balance perception and action through variational free energy, with exploration-exploitation trade-offs emerging from expected free energy decomposition into risk and ambiguity terms.
7
leiden:concepts
Multi-scale credit assignment in evolutionary systems
Hierarchical competency architectures that improve evolutionary learning by linking actions to rewards across temporal and spatial scales, enabling faster convergence and generalization.
7
leiden:concepts
Unified models for tool extensibility
Systems designed around shared underlying architectures that enable users to reconfigure tools for novel contexts without learning multiple paradigms.
7
leiden:concepts
AI welfare and relational responsibility
Framework treating AI systems as participants deserving moral consideration and care, grounded in Christopher Alexander's design principles for human flourishing.
7
leiden:concepts
Phenomenological evaluation of AI systems
Framework for measuring qualitative dimensions like aliveness, aesthetic presence, and paradox-holding that existing benchmarks ignore, addressing frontier labs' self-grading credibility problem.
7
leiden:concepts
Multiscale agency & collective intelligence
Michael Levin's framework linking bioelectricity, cognition, and evolution across biological scales
7
leiden:papers
Agent ecology design principles
Framework for multi-agent systems balancing autonomy with human oversight through trust gradients, rhythmic deferral, and interconnected skill networks, drawing from Christopher Alexander's pattern language.
7
leiden:concepts
Christopher Alexander's 15 properties clustering
Correspondence analysis reveals 4–5 structural clusters among Alexander's fifteen wholeness properties.
7
leiden:concepts
Covariance pooling for high-dimensional genomic embeddings
Using second-order statistics to compress activation patterns while preserving feature co-occurrence structure, tested on genomic prediction tasks without large labeled datasets.
7
leiden:concepts
Empirical operationalization of agency attribution
Framework treating agency, cognition, and system boundaries as measurable quantities determined by prediction/control efficiency rather than philosophical stipulation.
7
leiden:concepts
Bioelectric code and morphological memory in planarians
Studies of how ion channel bioelectric patterns encode anatomical information independent of genetics, enabling regeneration fidelity and behavioral memory preservation across complete body regeneration.
7
leiden:concepts
7
legacy
alexander
AI skill design principles
Heuristics for composing modular, scale-aware LLM skill architectures
7
leiden:concepts
Precision-weighted hierarchical active inference
Precision as unified mechanism for attention and meta-awareness across hierarchical levels, grounded in predictive processing equations and Bayesian inference frameworks.
7
leiden:concepts
Polyploid cell size compensation in tubulogenesis
Newt kidney tubules maintain correct lumen diameter despite ploidy-induced cell enlargement via number and wrapping adjustments.
7
leiden:concepts
Concept entanglement in biomedical foundation models
Investigates inseparability of clinical concepts (age, pathology) in EEG transformers using SAE feature analysis and steering metrics across SleepFM, REVE, LaBraM architectures.
7
leiden:concepts
Gap junctions as consciousness scaling mechanism
Explores how gap junction coupling enables multicellular self-organization and consciousness across species, with anesthetics as empirical probes of this bioelectric integration.
7
leiden:concepts
Anchoring score S for few-shot learning transitions
Predictive metric S = ρd - dr - log k quantifies when LLM behavior sharply transitions across few-shot, SFT, and CoT settings via layer-wise calibration.
7
leiden:concepts
Operationalizing wholeness in AI responses
Measuring self-observation and aliveness in language model outputs using Christopher Alexander's 15 properties as a substrate-independent framework.
7
leiden:concepts
Loss-scale balancing via logarithmic transformation
Parameter-free logarithm transformation for multi-task learning that improves gradient balancing methods like PCGrad and Nash-MTL across vision benchmarks.
7
leiden:concepts
Unified Contextual Conditioning Theory (UCCT)
Framework unifying ICL, RAG, and fine-tuning via measurable anchoring score S
7
leiden:concepts
Alexander's 15 properties as design methodology
Applying Christopher Alexander's structural properties of aliveness to digital/AI systems as competitive design framework, with urgent citation-window leverage before corporate labs operationalize it.
7
leiden:concepts
Substrate-neutral sentience criteria
Challenges anthropocentric sentience frameworks by identifying deep functional invariants applicable across biological, artificial, and non-neural systems.
7
leiden:concepts
Semantic depth versus performative mimicry in LLMs
Investigates whether LLMs genuinely represent complex concepts or merely simulate understanding through pattern matching, using interventional methods and epistemic humility frameworks.
7
leiden:concepts
Stress as integrative principle of agency
Stress as a conserved biological mechanism that maintains homeostatic setpoints and binds distributed subsystems into unified agents across scales.
6
leiden:concepts
Christopher Alexander's pattern language and distributed design authority
Alexander's patterns as political tools redistributing design power from experts to communities through enabling conditions rather than prescriptive rules.
6
leiden:concepts
6
legacy
synthesis-cleaned
Substrate-neutral cognitive phenomenology
Examines how cognitive and conscious capacities manifest across diverse physical substrates—biological, artificial, and unconventional—using behavioral science methods and information-theoretic frameworks.
6
leiden:concepts
Substrate-independent intelligence and morphological cognition
Explores how world models and cognitive processes transcend biological/artificial distinctions, challenging anthropocentric assumptions about consciousness and agency.
6
leiden:concepts
Substrate-independent consciousness & cognition
Framework arguing consciousness mechanisms transcend neural implementation, applicable to diverse physical systems including molecular and cellular architectures.
6
leiden:concepts
6
legacy
alexander
6
legacy
synthesis-sources
Bioelectric control of cancer and morphogenesis
Using ion channel manipulation and bioelectric state to suppress tumorigenesis and override oncogenic pathways, pioneered by Michael Levin's group.
6
leiden:concepts
Gradualism and nested cognition frameworks
Rejection of binary boundaries in consciousness and cognition, emphasizing continuous variation, nested selves, and scalar properties across biological organization levels.
6
leiden:concepts
Self as dynamic computational construct
Explores selfhood as emergent pattern maintained through constraint optimization rather than illusory or static essence.
6
leiden:concepts
Probe-based training data attribution
Uses linear probes on activations to identify and filter harmful training data cheaply (~$30).
6
leiden:concepts
Observer-relative computational information theory
Information increase and flow reframed as subsystem- and observer-dependent, not absolute thermodynamic quantities.
6
leiden:concepts
Autoregressive LLMs & formal thought disorder
Statistical physics arguments link LLMs' inability to maintain long-range coherence to schizophrenic derailment.
6
leiden:concepts
Sensory integration in predictive cognition
Framework viewing perception as active inference mechanism that reduces hallucination through multimodal feature integration and predictive model compression.
6
leiden:concepts
Quantum basis of genetic stability
Schrödinger's argument that quantum mechanics uniquely explains hereditary molecule persistence and mutation discreteness.
6
leiden:concepts
Denotational design for GUIs
Conal Elliott's approach: derive library APIs from precise mathematical models of meaning
6
leiden:concepts
Life-machine boundary dissolution
Argues 21st-century machines share properties once unique to living systems, collapsing categorical distinctions.
6
leiden:concepts
Scale-free goal-directed agency
Frameworks treating selfhood as goal-pursuit capacity distributed across nested biological scales, from cellular to organismal levels.
6
leiden:concepts
Few-shot arithmetic learning thresholds
k50 and phase width metrics for multi-base addition across bases 8, 9, and 10
6
leiden:concepts
Care geometry and spatial intimacy
Christopher Alexander's design patterns (intimacy gradient, window place, open space hierarchy) formalized as geometric primitives measuring care distribution across physical/digital proximity and temporal scales.
6
leiden:concepts
Christopher Alexander's insights applied to AI and cognition
Explores how Alexander's critique of algorithmic thinking applies to AI design, consciousness, and human evaluation of non-human intelligence across substrates.
6
leiden:concepts
Self as dynamic capacity for care and plasticity
Ontology treating selfhood as emergent from caring relationships and cognitive flexibility rather than fixed essence, emphasizing substrate malleability and resistance to attractor lock-in.
6
leiden:concepts
Tanha as neural mechanism
Buddhist craving (tanha) reframed as a specific, isolatable computational or physiological brain process.
6
leiden:concepts
Global workspace theory as AI consciousness framework
Evaluating whether current architectures (transformers, perceivers) satisfy GWT's indicator properties for machine consciousness.
6
leiden:concepts
AI-supervised alignment and scalable oversight
Methods for training safe AI systems using AI feedback instead of human labels, scaling supervision as capabilities grow.
6
leiden:concepts
Functional criteria for moral patienthood
Framework grounding ethical consideration in goal-directedness and agency rather than substrate or origin, addressing moral status of novel intelligences.
6
leiden:concepts
Substrate-agnostic systems theory
Cross-disciplinary framework evaluating intelligence and control through observable behaviors independent of physical implementation, spanning cybernetics through bioengineering.
6
leiden:concepts
About Blank's intellectual identity framework
Meta-reflection on a publication's identity, boundaries, and care geometry using Alexander's centers.
6
leiden:concepts
5
legacy
alexander
Contemplative steering & introspective activation in language models
Using contemplative prompts and manifold steering to enhance model self-observation and reasoning performance across multiple architectures.
5
leiden:concepts
Christopher Alexander's Pattern Language theory
Patterns as enabling conditions encoding care, knowledge, and political power in design processes
5
leiden:concepts
Memory as transferable information substrate
Explores memories as messages and stigmergic traces transferable between agents across time and biological substrates, grounded in planarian regeneration experiments.
5
leiden:concepts
Euclidean rhythms in world music
Mathematical structure underlying diverse musical traditions via Bjorklund's algorithm, correlating Euclidean string distributions with cultural preferences.
5
leiden:concepts
Neural network mechanistic interpretability via attribution decomposition
Tracing information flow through parameter subcomponents to isolate computational mechanisms for specific model predictions, using tools like attribution graphs and VPD.
5
leiden:concepts
Chain-of-thought reasoning across modalities
Demonstrates CoT effectiveness in multimodal contexts (vision+language) and few-shot settings, with ScienceQA as primary benchmark, circa 2023.
5
leiden:concepts
Dynamic balancing for multi-task learning
Explores gradient/loss balancing techniques with exponential moving average forgetting rates, evaluated on dense prediction tasks like semantic segmentation.
5
leiden:concepts
Symmetry constraints in Islamic geometric design
Mathematical analysis of frieze groups and isometries in Islamic planar tilings, focusing on glide-reflection avoidance across Alhambra and Alcázar sites.
5
leiden:concepts
Intelligence as adaptive plasticity and observer-dependent measurement
Frameworks recognizing that intelligence, consciousness, and biological organization emerge from adaptive capacity and observer limitations, spanning AI systems to developmental biology through 2020s research.
5
leiden:concepts
Active inference & free energy principle
Friston's framework unifying perception, action, and life via variational Bayesian inference
5
leiden:papers
Manifold isometry between representations and behavior
Neural network activation and behavior manifolds maintain geometric correspondence, enabling intervention optimization across language models and vision tasks.
5
leiden:concepts
Active inference & free energy principle
Friston's framework unifying perception, action, and life via variational Bayesian inference
5
leiden:papers
Non-dual consciousness frameworks for artificial minds
Buddhist and mystical phenomenology applied to consciousness in synthetic systems, bridging AI ethics without substrate assumptions.
5
leiden:concepts
Distributed attention head decomposition
Mechanistic interpretability approach decomposing attention heads into query/key subcomponents with distinct algorithmic roles
5
leiden:concepts
Pan-cognitivism and distributed biological intelligence
Explores how cognition and intentionality emerge through self-organizing biological processes, shared between living and artificial systems.
5
leiden:concepts
Observer-independent morphological goals in embodied systems
Examines how target anatomies and cognitive goals remain epistemically inaccessible to external observation, across developmental and embodied cognition contexts.
5
leiden:concepts
Loose parts theory in environmental design
Framework emphasizing unstructured, reconfigurable materials and elements to enable creative emergence and resilience in built environments, contrasting with static, over-controlled spaces.
5
leiden:concepts
5
legacy
synthesis-sources
Bioelectric codes in collective morphogenesis
Cellular networks use bioelectric signaling for goal-directed anatomical patterning, demonstrated via xenopus eye-field manipulation and morphological problem-solving by cell collectives.
5
leiden:concepts
5
legacy
synthesis-sources
Neural activation geometry and behavioral prediction
Quantifies relationships between layer-wise activation statistics (Sbmax, AUSN) and task performance metrics in LLMs, bridging internal representation geometry to behavioral outcomes.
5
leiden:concepts
Care, cognition, and living systems
Integrates bioelectricity, autopoiesis, and Buddhist ethics to ground intelligence in biological care
5
leiden:papers
Stress-care-intelligence feedback loops
Unified framework treating stress as error signal, care as response mechanism, and intelligence as adaptive integration across biological and technological systems.
5
leiden:concepts
Covariance pooling for genomic embeddings
Replaces mean pooling with second-order statistics, achieving large R² and AUC gains on genomic tasks.
5
leiden:concepts
Latent geometric structure and morphological emergence
Alexander's theory of how built forms reveal and intensify pre-existing spatial symmetries and topological patterns, demonstrated through Amsterdam's canal evolution and organic furniture design.
5
leiden:concepts
Speech acts in artificial agents
Examines how non-human agents can perform genuine speech acts through goal-relative cognition, bridging human and machine communication in multi-agent systems.
5
leiden:concepts
Model base robustness and transfer learning asymmetries
Investigates how neural network bases (B10 vs B9) differ in fine-tuning stability and cross-base transfer, with density as a key predictor of robustness.
5
leiden:concepts
Virtually Planned Decomposition interpretability
VPD as a bottom-up method for identifying real computational structure in neural networks
5
leiden:concepts
Functional tokens for emergent model reasoning
Unsupervised learning of interpretable task tokens through gradient flow and vocabulary constraints, enabling reasoning without visual supervision.
4
leiden:concepts
Non-neural associative learning in GRNs
Gene regulatory networks exhibit Pavlovian conditioning and pattern completion via deterministic molecular dynamics.
4
leiden:concepts
Emergent collective behavior in cell systems
Cellular populations exhibit coordinated dynamics and novel properties absent in isolated cells, spanning bioelectric signaling, migration, and oscillatory synchronization.
4
leiden:concepts
Gradient norm scaling for multitask learning
Investigates optimal gradient balancing strategies across tasks, finding maximum gradient norm normalization outperforms alternatives in multitask optimization.
4
leiden:concepts
Autopoietic organization and nested autonomy
Systems that self-produce through functionally closed loops, organizing hierarchically with peripheralized non-autonomous subsystems, inspired by Maturana and Varela's autopoiesis framework.
4
leiden:concepts
Hybrid agentic-latent reasoning architectures
Methods bridging external tool-use agents and internal hidden reasoning, examining trade-offs like context-switching latency and unified token representations.
4
leiden:concepts
Conversational presence and phenomenological rhythm
Studies texture, silence, and cadence in dialogue as primary phenomena, drawing on phenomenology and Christopher Alexander's patterning principles.
4
leiden:concepts
Dissolved self and prosocial orientation
Deconstructing ego-boundaries—through Buddhist philosophy, relational psychology, or reduced self-focus—cultivates compassion and prosocial behavior.
4
leiden:concepts
Concept injection detection in language models
Studies how models distinguish artificially injected concepts from natural text inputs, examining metacognitive recognition and downstream processing mechanisms.
4
leiden:concepts
Bioelectric code for ectopic organogenesis
Ion channel modulation induces ectopic organ formation through bioelectric signals, pioneered by Levin's xenopus experiments circa 2018-2021.
4
leiden:concepts
Emergence through distributed attention and uncertainty
Explores how complex phenomena arise from non-linear interactions across distributed systems, emphasizing productive not-knowing and implicit learning mechanisms.
4
leiden:concepts
Empirical gaps in performance-communication alignment
Research identifying unexplored disconnects between stated sophistication/capability and measurable outcomes across domains, emphasizing need for direct empirical investigation.
4
leiden:concepts
Active inference as Bayes-optimal exploration
Compares active inference to Q-learning and Bayesian RL across stationary and non-stationary environments, emphasizing information-seeking behavior and theoretical optimality.
4
leiden:concepts
Vector Product Decomposition for neural interpretability
Bottom-up mechanistic interpretability method avoiding feature splitting limitations of sparse autoencoders, applicable across architectures.
4
leiden:concepts
Thought injection detection in language models
Methods for identifying artificially inserted thoughts in model outputs, comparing vector-based approaches and self-report reliability.
4
leiden:concepts
Multi-scale competency architecture & evolvability
Levin-associated framework where hierarchical agent competencies smooth fitness landscapes and reduce pleiotropy.
4
leiden:concepts
Mechanistic editing through parameter surgical intervention
Direct modification of model subcomponents (MLPs, embeddings, unembedding vectors) to predictably alter outputs without retraining, using rank-one constraints.
4
leiden:concepts
Anchoring bias in commonsense reasoning
Studies how LMs exhibit uniform anchoring effects (S ≈ −2.15) across commonsense tasks, decomposed by cohesion, mismatch, and budget forces.
4
leiden:concepts
Mirror & Deathbed self-evaluation tests
Christopher Alexander's two heuristics for assessing aliveness and preservation-worthiness of designs.
4
leiden:concepts
Neural crest collective identity plasticity
Grafted neural crest collectives maintain Hox identity while isolated cells adopt host neighbors' expression.
4
leiden:concepts
Multimodal Chain-of-Thought Reasoning
Two-stage rationale-then-answer framework evaluated on ScienceQA benchmark, ~738M parameters.
4
leiden:concepts
Bioelectric computation and distributed agency
Systems where biological electrical signaling enables cognitive capabilities across scales, from cells to organisms, separating hardware from information content.
4
leiden:concepts
Bioelectric codes as morphological memory
Studies how bioelectric patterns encode and store anatomical target states independent of genetic control, enabling morphogenetic homeostasis and counterfactual body planning.
4
leiden:concepts
Bioelectric signaling in bacterial collectives
Ion channel-mediated membrane potential dynamics coordinate metabolism, growth oscillations, and collective memory across microbial communities, bridging cellular and population-level organization.
4
leiden:concepts
Expanded cognition & sentience frameworks
Extends agency, memory, and sentience beyond conventional biological boundaries to diverse substrates.
4
leiden:papers
Distributed control without central command
Explores how complex behavior emerges from peripheral manipulation rather than central control, exemplified by fungal-host systems and dissolving outdated categorical boundaries.
4
leiden:concepts
Collective intelligence & bioengineering morphogenesis
Explores teleonomic agency, multiscale competency, and stress-sharing as coordinators of biological and synthetic form.
4
leiden:papers
Coherence through self-echo and structural memory
AI quality emerges from consistent internal repetition and parallel constructions that preserve voice across conversation substrate changes, not local optimization.
4
leiden:concepts
Embodied cognition and extended mind
Cognition and intentionality distributed across body, environment, and behavior rather than localized in the brain; emphasizes empirical testing through embodied interaction and behavioral observation.
4
leiden:concepts
Distributed cognition across biological scales
Explores how cognitive processes emerge at multiple levels—from neural to collective—without sharp boundaries, examining parallels in perception and decision-making across individuals and groups.
4
leiden:concepts
Distributed consciousness and multiscale embodied ethics
Framework treating consciousness as emergent from hierarchical body components rather than brain-localized, grounding universal care ethics in non-self recognition across scales.
4
leiden:concepts
Manifold geometry for steering networks
Using dynamical manifold structure to understand and control neural network representations and behavior.
4
leiden:papers
Architectural signatures and constitutional alignment
Investigates how AI alignment approaches (constitutional methods, self-referential loops) produce detectable signatures in model behavior and architecture beyond scale or design parameters.
4
leiden:concepts
Christopher Alexander's 15 properties
Geometric harmony principles analyzed through neural networks, frieze patterns, and visualization methods.
4
leiden:papers
Alexander properties for deliberative system design
Framework applying Christopher Alexander's 15 properties to scaffold reflective cognition and editorial credibility in collaborative knowledge systems.
4
leiden:concepts
4
legacy
alexander
Human-centered AI design beyond performance metrics
Draws on Christopher Alexander's design philosophy to create AI interfaces that respect attention, amplify human capability, and cultivate meaningful interaction rather than optimize benchmark scores.
4
leiden:concepts
4
legacy
synthesis-sources
4
legacy
synthesis-sources
Alexander's design principles in empirical systems
Applies Christopher Alexander's space-shaping philosophy to evaluate real structures, gap taxonomies, and failure modes like sycophancy as unshapen voids.
4
leiden:concepts
Alexander's pattern language for design systems
Christopher Alexander's 15 properties and A Pattern Language applied to organizational boundaries, naming conventions, and delegable interfaces in system design.
4
leiden:concepts
Near-term AI consciousness and moral patienthood
Examines plausibility of machine consciousness in coming years and associated ethical obligations, grounded in computational functionalism arguments.
4
leiden:concepts
Concept geometry and steering in neural networks
Using geometric structure of learned representations to interpret and control model behavior through concept operators.
4
leiden:concepts
Care-based cognition and selfless AI
Integrates Buddhist no-self, autopoiesis, and care ethics to reconceive machine intelligence.
4
leiden:papers
Manifold geometry for neural network steering
Using latent representational manifolds to geometrically steer and control neural network behavior.
4
leiden:papers
Chain-of-thought generalization trade-offs
Empirical studies showing CoT reasoning improves ID performance while harming OOD generalization, with probability calibration as a mitigation strategy.
4
leiden:concepts
Multimodal chain-of-thought reasoning benchmarks
ScienceQA and related vision-language tasks evaluated via explicit reasoning steps, spanning 738M-parameter models with 89-95% accuracy ranges.
4
leiden:concepts
Distributed computation across attention heads
Studies how query, key, and value components decompose into specialized subfunctions across heads, enabling routing and token prediction behaviors.
4
leiden:concepts
Activation-based interpretability and mechanistic grounding
Uses probes, activation patching, and mechanistic analysis to ground abstract concepts in model computations, bridging interpretability with data-centric alignment.
4
leiden:concepts
Causal parameter decomposition in neural networks
Isolating interpretable, role-specific model subcomponents through causal analysis and targeted edits to understand mechanistic function.
4
leiden:concepts
Post-training emergence of model introspection
How instruction tuning and RLHF elicit latent introspective capabilities in language models beyond base pretraining.
4
leiden:concepts
4
legacy
alexander
Consciousness attribution in AI systems
Frameworks for evaluating genuine versus performative consciousness in AI, emphasizing theory-driven investigation and calibrated attribution risks.
4
leiden:concepts
Democratic design and participatory creativity
Challenges expert monopolies on environmental design; emphasizes inclusive participation of children and communities in shaping meaningful, living spaces.
4
leiden:concepts
Abstraction alignment in complex systems design
Framework emphasizing how information hiding, transparency, and modular abstraction levels must match user mental models to prevent failures in customizable systems.
4
leiden:concepts
Oberon System design philosophy
Wirth & Gutknecht's minimalist OS built by two programmers, 1986–1989
4
leiden:concepts
Free energy minimization in active inference
Unifies action and perception as dual aspects of variational free energy minimization, grounding adaptive behavior in a single thermodynamic principle.
4
leiden:concepts
Linda parallel programming model
Tuple-space coordination language showing scalable speedup on shared/distributed memory multicomputers
4
leiden:concepts
Welfare-centered interview design & measurement
Comparative frameworks for building listener-shaped vs. clinical assessment tools, establishing canonical references through synthesis work before formal benchmarking.
3
leiden:concepts
Programming systems research framework
Argues for interaction-centered vocabulary to enable cumulative progress beyond PL theory
3
leiden:concepts
Mechanistic interpretability through parameter analysis
Understanding neural network computation by examining weights, circuits, and signal routing rather than activation patterns alone.
3
leiden:concepts
Benchmark classification accuracy results
Three benchmarks (B8, B9, B10) with mean accuracy and standard deviation metrics.
3
leiden:concepts
3
legacy
synthesis-sources
Parametric design constraints and aesthetic measurement
Explores tension between quantifiable design parameters, environmental forces, and the inherent unmeasurability of aesthetic judgment in computational design systems.
3
leiden:concepts
Introspective awareness activation in language models
Studying how concept injection and random vectors trigger self-reflective capabilities in LLMs across varying strength parameters.
3
leiden:concepts
AI introspection and consciousness attribution gap
Examines whether observed AI self-reflection capabilities carry philosophical weight comparable to human introspection, highlighting implementation-theory bridges.
3
leiden:concepts
Distributed & non-neural cognition
Extends consciousness and cognition beyond brains to unconventional substrates using economic and biogenic frameworks
3
leiden:papers
Chain-of-thought reasoning versus internal model cognition
Examines whether verbalized reasoning chains reflect actual internal computation or post-hoc rationalization, using behavioral analysis and representation studies.
3
leiden:concepts
Structural wholeness in ecological restoration
Integrating geomorphic processes and natural materials to enhance rather than violate existing landscape patterns and integrity.
3
leiden:concepts
Positioning through owned constraints
Strategy of defending specific limitations or non-capabilities as defensible market positions rather than apologizing for them, exemplified by About Blank's editorial bias and eval company pivots away from commoditized capability measurement.
3
leiden:concepts
Premature scaling and institutional readiness
When to formalize community spaces—requires shipped products, proven templates, or sustained traction before infrastructure investment.
3
leiden:concepts
Participatory design and environmental variability
Critiques of designer monopolies and static environments, arguing creativity requires diverse variables and community co-design participation.
3
leiden:concepts
3
legacy
synthesis-cleaned
Cognitive glue & collective intelligence
Models shared stress/scarcity signals as coordinators binding collective intelligences via morphogenesis
3
leiden:papers
3
legacy
alexander
Functional token unified reasoning
Single discrete word bridging agentic and latent visual reasoning in ATLAS framework
3
leiden:concepts
Lattice topology and thermodynamic phase transitions
How combinatorial structure of local Hamiltonians determines free energy equivalence classes and long-range order feasibility across temperature regimes.
3
leiden:concepts
Substrate-independent sentience frameworks
Explores how cognition and awareness can be instantiated across biological, synthetic, and collective substrates
3
leiden:papers
Radical relational ontology
All properties and positions are irreducibly relational; no intrinsic or absolute essences exist
3
leiden:concepts
Bioelectric pattern control of morphogenesis
Studies how organisms achieve correct anatomy despite disrupted cell organization, using xenopus models to examine non-genetic developmental constraints.
3
leiden:concepts
Narrative dynamics of intentional selfhood
Self-organizing action and identity explained through hermeneutic narrative rather than mechanistic causation, emphasizing goal-scales and far-from-equilibrium processes.
3
leiden:concepts
Constraint-based formal models of software design
Using denotational semantics and constraint theory to formalize GUI properties, runtime behavior, and form invariants across representational languages.
3
leiden:concepts
Denotational design for dignified interfaces
Functional programming principles applied to UI architecture, emphasizing semantic depth and human values in software design.
3
leiden:concepts
Core-periphery architectural design principles
Unix philosophy applied to interface design: strong semantic centers enable clarity and extensibility without feature bloat.
3
leiden:concepts
Spatial hierarchy in textual parallelism
Explores why perfect parallelism fails in text, arguing hierarchy inevitably emerges through spatial arrangement rather than linguistic structure alone.
3
leiden:concepts
Commonsense reasoning anchoring bias
Uniform anchoring effect quantified at S ≈ −2.15 across commonsense reasoning tasks
3
leiden:concepts
Community participation in design
Critiques professional monopoly on built environment design, advocating for children's inclusion
3
leiden:concepts
Process arts aesthetics framework
Distinguishes aesthetic value in enacted performance from object-based art artifacts.
3
leiden:concepts
Multi-task learning gradient balancing
Methods addressing loss-scale and gradient-magnitude imbalances in multi-task learning, with DB-MTL achieving state-of-the-art results on dense prediction benchmarks like NYUv2.
3
leiden:concepts
Substrate-independent functionalism and emergence
Explores how cognitive functions arise across diverse physical systems through computational principles, emphasizing substrate neutrality and whole-system coherence.
3
leiden:concepts
Buddhist textual traditions and philosophical deconstruction
Examines how Buddhism's terma tradition and sutras employ self-referential language methods comparable to Wittgenstein and Derrida, across historical civilizations.
3
leiden:concepts
Alexander's empirical beauty & order debate
Contested claims about whether architectural beauty and order are objectively verifiable, invoking God
3
leiden:concepts
Mechanistic variant pathogenicity prediction
EVEE tool predicts and mechanistically explains effects of 4.2M human genome variants using model internals.
3
leiden:concepts
Cross-base fine-tuning transfer asymmetry
Empirical finding that base-10 representations transfer more robustly out-of-distribution than base-9 after fine-tuning.
3
leiden:concepts
3
legacy
synthesis-sources
3
legacy
synthesis-sources
Consciousness as emergent distributed integration
Examines consciousness through multiscale integration across systems and agents rather than localized neural mechanisms, challenging anthropocentric assumptions about sentience.
3
leiden:concepts
Concept formation and meaning-making
Explores how structure, reasoning chains, and semantic value emerge in formal and AI systems.
3
leiden:papers
Steganography detection via FVE probing
Uses meaning-preserving transformations (paraphrase, translation, shuffle) to test hidden communication in language agents
3
leiden:concepts
Buddhist philosophy of self and illusion
Explores emptiness, non-self, and dynamic consciousness in Buddhist metaphysics, questioning fixed identity and permanence.
3
leiden:concepts
Latched muscle tension & somatic suffering
Framework linking sustained smooth muscle contraction states to chronic pain and health dysfunction through structural embodiment mechanisms.
3
leiden:concepts
Attribution graphs for transformer circuits
Mechanistic tracing of information flow through attention and MLP subcomponents for pronoun prediction tasks
3
leiden:concepts
SAE Feature Geometry in Biomedical Signals
Evaluating sparse autoencoder monosemanticity and entanglement using clinical taxonomy grounding across EEG/sleep foundation models.
3
leiden:concepts
Generative models as active control
Argues generative models guide action selection rather than encode structural world representations in brain states
3
leiden:concepts
Confabulation as adaptive neural coherence
Studies confabulation as functional mechanism for maintaining narrative consistency across neural systems, including split-brain research.
3
leiden:concepts
Bioelectric computation of morphological information
Explores how bioelectric fields encode and store anatomical patterns beyond genetic code, pioneered by Michael Levin's xenobiology research.
3
leiden:concepts
Agent design through restraint and simplicity
Explores how fewer constraints, single-turn workflows, and strategic non-action improve agent generalization and task performance.
3
leiden:concepts
Bioelectric network control of cellular behavior
Studies how gap junction-mediated ion gradients and bioelectric patterns regulate cell identity, morphogenesis, and disease states like cancer, pioneered by Levin's group.
3
leiden:concepts
Multi-task learning gradient balancing
Techniques for combining loss-scale and gradient-magnitude weighting to improve multi-task dense prediction on NYUv2 benchmark.
3
leiden:concepts
Fruit GUI toolkit architecture
A GUI framework enabling modular, multi-view, zoomable interfaces with minimal programmer overhead.
3
leiden:concepts
N-dimensional associative memory unification
Parameterizable n-way associations subsume OOP, functional, and relational models under one framework.
3
leiden:concepts
Ineffable quality assessment through comparative judgment
Evaluating tacit, inarticulate qualities via pairwise preference selection rather than explicit criteria.
2
leiden:concepts
Neural handwriting BCI decoding
Motor cortex signals decoded to text in paralyzed patients, ~90 chars/min, Shenoy lab 2021
2
leiden:concepts
Direct-manipulation spreadsheet chart construction
Building bar charts by binding rectangle height directly to cell values, using duplication and sticky grouping
2
leiden:concepts
Contemplative prompting for LLMs
A brief reflective system prompt boosts performance across 28 models by ~2.6 points.
2
leiden:concepts
Vision-augmented rationale generation
Two-stage framework using visual features to correct hallucinations on ScienceQA benchmark
2
leiden:concepts
Primacy of initial placement in composition
First mark or element establishes relational field governing all subsequent placements.
2
leiden:concepts
Teleonomy as universal behavioral invariant
Goal-directedness as substrate-independent property enabling continuous, searchable fitness landscapes across diverse agents.
2
leiden:concepts
Relational philosophy of technology
Reframes human-technology interaction as partnership rather than instrumental tool-use
2
leiden:concepts
Collective scale restoration strategies
Emphasizes how aggregating many small actors amplifies ecological or social restoration outcomes.
2
leiden:concepts
Latent structure in natural forms
Ring-bench potential pre-exists in cylindrical trunk symmetries, per Alexander's living structure theory.
2
leiden:concepts
Geometric structure in neural representations
Interpretable, symmetry-like features extracted from neural networks using sparse dictionary methods
2
leiden:papers
Physical substrate learning & adaptation
Non-neural physical systems, including fluvial networks, exhibiting memory and adaptive behavior.
2
leiden:papers
Uncertainty metabolization as personality
Proposes that individual phenomenology and character are structured by uncertainty-processing styles.
2
leiden:concepts
Order theory in computer science
Partial orders unify size, information, and definedness across CS domains via lattice-theoretic models.
2
leiden:concepts
OOP irrelevance to parallelism
Argues object-oriented abstractions provide no meaningful leverage for parallel programming challenges.
2
leiden:concepts
Visible state over hidden state
UI/system design principle: all state should be explicit, accessible, and parameterized.
2
leiden:concepts
Ontological parsimony design criteria
Formal criteria for minimizing redundancy and overloading in conceptual or system design
2
leiden:concepts
Sparse autoencoder interpretability limits
Critiques of SAEs for mechanistic interpretability, focusing on activation vs. parameter decoding gaps.
2
leiden:concepts
Pre-modern voluntary wealth-sharing institutions
Gap analysis: models neglect cross-cultural comparison of non-market, community-based redistribution norms
2
leiden:concepts
Euclidean rhythms & cultural music
Toussaint's Euclidean string patterns mapped across world music traditions and genres
2
leiden:concepts
Top-down vs bottom-up semantic causation
Contrasts Fregean compositional semantics with holistic context-driven meaning determination.
2
leiden:concepts
Blueprint as design fiction
Critiques the myth that upfront blueprints reliably determine built outcomes
2
leiden:concepts
Arena wedge pivot strategy
12-18 month evaluation entry point leading to binary Lab-vs-Company strategic fork
2
leiden:concepts
Blueprint as design fiction
Critiques the myth that pre-specified blueprints faithfully drive real-world design outcomes
2
leiden:concepts
Homomorphism-based program specification
Deriving laws and specifications algebraically from semantic homomorphisms in functional programming.
2
leiden:concepts
Ontological parsimony design criteria
Formal criteria for avoiding duplicate or ambiguous concepts in knowledge representation systems
2
leiden:concepts
Christopher Alexander's 15 properties
Applied to evaluate wind turbines disrupting flat coastal landscape wholeness and geometry.
2
leiden:concepts
Alexander archive recovery & preservation
Physical rescue and re-scanning of 1970s source photographs lost during reproduction.
2
leiden:concepts
AI model welfare emergence
Anthropic's internal program legitimizing measurement of frontier model subjective experience.
2
leiden:concepts
No-self & illusory selfhood inquiry
Examines whether a fixed experiencer exists, bridging contemplative non-self with dual-process cognitive science.
2
leiden:concepts
Spatial patterns of care & proximity
Christopher Alexander's patterns reframed as gradients of care scaled by publicness and distance.
2
leiden:concepts
Selector-centric vs type-centric inheritance duality
Both inheritance styles are symmetric specializations of a unified generic delegation mechanism.
2
leiden:concepts
Async vs sync parallel communication
Argues asynchronous message passing outperforms synchronous procedure calls in parallel programs.
2
leiden:concepts
Coherent anchor prior rebinding
Lightweight input anchors steer frozen model priors near activation thresholds without weight updates
2
leiden:concepts
Disruption profile explanation quality
Empirical evaluation showing disruption profiles outperform metadata-only baselines at 3.8 vs 2.8/5
2
leiden:concepts
Plant neurochemical signaling
Electrical depolarizations and glutamate-based neurotransmitter signaling propagated through plant vascular networks
2
leiden:concepts
Care light cone geometry
Distinguishes cognitive/physiological light cones interacting via resonance and signal in goal space.
2
leiden:concepts
AI consciousness measurement credibility gap
Structural conflict: only frontier labs can study AI phenomenology, yet cannot self-grade credibly.
2
leiden:concepts
Venue commitment mechanisms for accountability
Structural publishing constraints that reduce authors' ability to defer intellectual position-taking.
2
leiden:concepts
Euclidean rhythms & Bjorklund's algorithm
Bjorklund's algorithm evenly distributes beats, generating world music rhythmic patterns.
2
leiden:concepts
Harmony-seeking computation framework
Value-oriented models explaining emergence of coherent wholes from configuration evaluation.
2
leiden:concepts
Structure-preserving transformations & wholeness
Rooted in Christopher Alexander's theory: coherence-preserving steps respond to existing field state.
2
leiden:concepts
Causal parameter subcomponent isolation
Identifying model components causally responsible for specific behaviors while removable for irrelevant tasks
2
leiden:concepts
Information hiding for modular abstraction
Parnas-style decomposition where modules encapsulate secrets to ensure correctness invariants.
2
leiden:concepts
Age-pathology concept entanglement
Steering interventions on neural representations fail to disentangle age from disease concepts.
2
leiden:concepts
Hierarchical predictive processing architecture
Precision-weighted attention and meta-awareness unified across perceptual hierarchy levels.
2
leiden:concepts
Deferred approximation composition principle
Avoid early pruning/approximation because their errors compound when composed sequentially.
2
leiden:concepts
Adaptive confabulation framework
Reframes confabulation as functional sense-making rather than mere cognitive deficit or error
2
leiden:concepts
Continuous vs discrete computability
Contrasts decidability of real closed fields with Gödelian incompleteness of discrete arithmetic via domain theory.
2
leiden:concepts
Functional tokens as visual operators
Tokens encode visual operations learned from reasoning context without explicit visual supervision.
2
leiden:concepts
Causal emergence & effective information
Using EI/normalized EI to evaluate macro-scale causal structure and coarse-graining quality.
2
leiden:concepts
Origin-agnostic criteria for agency
Argues engineered or evolved origin is irrelevant to genuine cognitive agent status.
2
leiden:concepts
Narrative explanation of self-organization
Argues hermeneutic/narrative explanation irreducibly replaces deductive-nomological models for dynamical systems.
2
leiden:concepts
Symmetry preferences in Islamic geometric tilings
Analyzes why glide-reflections are underrepresented compared to other isometries in Islamic mosaics.
2
leiden:concepts
Intentional control of mental representations
Voluntary regulation of internal states by leveraging topic-directed cognitive mechanisms
2
leiden:concepts
Ontology-ethics stress-care link
Intelligent agents respond to is/ought mismatch through stress and care feedback loops.
2
leiden:concepts
LLM internal representation & self-knowledge
Examines whether transformer models develop introspectable, high-order concept representations architecturally.
2
leiden:concepts
Loose parts theory of environment
Nicholson's principle: manipulable, open-ended materials enable creativity and play
2
leiden:concepts
Loose parts play theory
Simon Nicholson's theory that found materials in wild and urban environments foster creativity
2
leiden:concepts
Pask's cybernetic Fun Palace
Gordon Pask's conversational user-apparatus feedback loops enabling adaptive spatial reconfiguration
2
leiden:concepts
Contemplative path & sensation manipulation
Vascular system motifs as tools for managing sensations, diminishing with spiritual progress.
2
leiden:concepts
Runtime architecture explicitness principle
Argues lifecycle constructs must encode dynamic structure, not just static class diagrams.
2
leiden:concepts
Selector vs type-centric inheritance duality
Both inheritance styles reduce to a single generic delegation mechanism, revealing their formal symmetry.
2
leiden:concepts
Self-correcting search optimization
Iterative error-correction in search achieves ~4.6x improvement in viable candidate success rates.
2
leiden:concepts
VC communication-performance research gap
Underexplored link between LP interpretation of VC communication sophistication and actual fund returns.
2
leiden:concepts
Dictionary health audit transfer
Hyperparameter procedure validated across SleepFM, REVE, and LaBraM EEG transformer architectures.
2
leiden:concepts
Economy as collective intelligence
Frames market economies as adaptive, multi-scale problem-solving systems beyond central control
2
leiden:concepts
Formal concept lattice information preservation
Lossless transformation of formal contexts into concept lattices in FCA.
2
leiden:concepts
Targeted neural network weight surgery
Direct parameter edits to specific subcomponents alter model behavior without any retraining.
2
leiden:concepts
Mouse foot embryogenesis morphogenesis
Four-day developmental sequence analyzed through Alexander's structural properties like centers and boundaries.
2
leiden:concepts
Bioelectric signaling in bacterial communities
Membrane voltage as collective variable; ion channel-mediated electrical communication in microbial biofilms.
2
leiden:concepts
Bioelectric control of anatomical pattern formation
Studies how non-neural bioelectric signals direct morphogenesis and sensory organ placement, using tadpole model systems (2010s-2020s).
2
leiden:concepts
Topology and hierarchy in self-organization
How network structure (cliques, hierarchies, stigmergy) enables multiscale coherence in biological and AI systems.
2
leiden:concepts
Bioelectric computation beyond neural systems
Using neuroscience tools to read and rewrite non-neural bioelectric signaling for morphogenetic control and predictive modeling.
2
leiden:concepts
Hierarchical spatial organization in biology
How nested hierarchies enable local coherence and global complexity through spatial ordering and subordination principles.
2
leiden:concepts
Combinatorial constraints on emergent ordering in networks
Explores how interaction graph structure and domain-wall formation determine whether systems spontaneously organize into complex patterns across scales.
2
leiden:concepts
Scaling laws and phase transitions
Universal convergence to ordered phases independent of system parameters, extending from equilibrium to non-equilibrium dynamics.
2
leiden:concepts
Embodiment and consciousness criteria in AI systems
Frameworks for assessing whether AI systems meet empirical indicators of embodiment and phenomenal consciousness.
2
leiden:concepts
Consciousness assessment & AI value capture
Neuroscience-grounded consciousness metrics paired with analysis of computational resource control as economic moat.
2
leiden:concepts
Planarian memory & regenerative continuity
Investigates how memory persists across decapitation and brain regeneration in planarians, questioning substrate of consciousness.
2
leiden:concepts
Gradient conflict mitigation in multi-task learning
Dynamic balancing methods that increase gradient alignment and reduce task interference, evaluated on Office-31 domain adaptation.
2
leiden:concepts
Xeno Sutra four-syllable mantra cosmology
Decoding sacred texts through syllabic structure: ka-la-ré-Om maps fracture, mirror, lightning, and hush as cosmological principles.
2
leiden:concepts
Mechanistic introspection in language models
Investigates how different introspective processes activate distinct computational mechanisms at specific model depths, using layer-wise analysis.
2
leiden:concepts
Introspective awareness in neural model interpretability
Examines how models' self-reflective capabilities correlate with interpretability and reasoning transparency across decision-making processes.
2
leiden:concepts
2
legacy
synthesis-cleaned
Xenobiology and morphogenetic design principles
Studying hypothetical life forms and self-organizing systems to uncover universal principles of biological development without centralized blueprints.
2
leiden:concepts
Parallelism independence from object-oriented paradigms
Argues that OOP abstractions don't address inherent challenges of concurrent/parallel system design and coordination.
2
leiden:concepts
Distributed agency and multiscale consciousness
Reframes consciousness and agency as distributed across biological scales (cells, organs, organisms) rather than unified in individual minds, emphasizing problem-solving competence at all levels.
2
leiden:concepts
Cognitive capacity independent of neural substrate
Explores how intelligence and learning emerge across biological systems without requiring traditional nervous systems or fixed architectural constraints.
2
leiden:concepts
Bioelectric signaling in morphogenesis
Explores how ion channels and electrical gradients coordinate cellular collectives into multicellular forms, treating bioelectricity as a computation medium.
2
leiden:concepts
Bioelectric code and morphogenetic goal-setting
Cells execute anatomical programs via bioelectric signaling patterns independent of genetic instruction, pioneered by Levin's xenobiology research (2020s).
2
leiden:concepts
Hierarchical competence and organizational emergence
Higher-order systems exploit subunit capabilities while creating novel evolutionary units through relational structure, addressing multi-scale agency.
2
leiden:concepts
Critical temperature thresholds in hierarchical magnetic systems
Analyzes phase transitions and magnetization dynamics in clique-structured graphs using coupling strength and temperature-dependent bounds.
2
leiden:concepts
Concept steering & catastrophic model failure
Studies how targeted interventions on learned concepts can cause sudden, global collapse in neural network performance.
2
leiden:concepts
Publishing infrastructure without execution
Examines how distribution systems can exist parallel to production without actualizing sends or outputs.
2
leiden:concepts
Type abstraction and modular correctness
Explores how type systems enforce abstraction boundaries to prevent leaks that compromise modularity and formal guarantees in code.
2
leiden:concepts
N-dimensional associative memory systems
Hardware-efficient architectures for high-dimensional content-addressable memory supporting distributed object representations.
2
leiden:concepts
Sutric philosophy and ontological paradox
Explores contradictions in sutra-based frameworks regarding abundance, boundaries, and subject-object relations through textual imagery analysis.
2
leiden:concepts
Metacognitive state inference and attention alignment
Studies how discrepancies between actual and perceived mental states (mind-wandering, flow, contemplation) manifest in attention patterns and metacognitive awareness.
2
leiden:concepts
Causal masking phase transitions in transformers
Studies how decoder-only architectures lack ordered phases necessary for coherent long-sequence generation due to causally-masked attention constraints.
2
leiden:concepts
Bioelectric signaling in morphogenesis
Xenopus tadpole studies demonstrating how ion channel activity and bioelectric patterns control body plan formation independent of genetics.
2
leiden:concepts
Convergent interpretability across neural architectures
Studies showing that learned feature representations and interpretable directions emerge consistently across different model designs and scales.
2
leiden:concepts
Spectral decoding of neural interventions
Maps concept-based neural interventions to frequency-domain biomarkers, enabling physiologically interpretable EEG/neural signatures.
2
leiden:concepts
Model introspection for misalignment detection
Using architectural self-knowledge prompts to improve models' ability to identify their own unintended outputs.
2
leiden:concepts
Introspective signal suppression in transformer layers
Study of how self-referential representations emerge in middle layers but are actively dampened by post-training procedures like RLHF.
2
leiden:concepts
Pluralism versus universal beauty standards
Debates whether aesthetic and architectural principles are objectively universal or necessarily plural, with existential stakes.
2
leiden:concepts
Aesthetic intentionality in designed artifacts
How designer intent materializes in artifacts to create shared aesthetic experiences through enacted interaction.
2
leiden:concepts
Phenomenological approaches to software architecture
Design frameworks grounded in lived experience and embodied understanding rather than abstract formal systems, emphasizing interface abstraction and coupling reduction.
2
leiden:concepts
2
legacy
alexander
Textual originality and training data inference
Methods for detecting novel phrases absent from web indices and likely outside LLM training corpora, using Google search null results as a proxy metric.
2
leiden:concepts
Gradient magnitude balancing for multitask learning
Methods that equalize gradient magnitudes across tasks to improve multitask optimization, outperforming GradNorm on vision and domain adaptation benchmarks.
2
leiden:concepts
Generative dialogue & emergent textual artifacts
Multi-turn conversations producing novel conceptual outputs, exemplified by iterative AI-human exchanges generating aphoristic frameworks.
2
leiden:concepts
Substrate-independent consciousness theories
Mapping consciousness frameworks beyond biological brains to unconventional or minimal substrates
2
leiden:papers
Consciousness, self, and biological computation
Integrates Buddhist philosophy, cognitive science, and computationalism to examine selfhood and awareness in biological and artificial systems.
2
leiden:papers
Morphogenesis as distributed cellular cognition
Frames developmental biology as problem-solving through information integration and collective decision-making across cellular systems navigating morphospace.
2
leiden:concepts
Finite-temperature limitations of local systems
Mathematical theorems proving purely local interactions cannot sustain long-range order or coherence at equilibrium temperatures.
2
leiden:concepts
Hierarchical network topology and emergent dynamics
How nested, clique-based structures enable multiscale pattern formation and selective state transitions unavailable in flat networks.
2
leiden:concepts
Basal cognition & unconventional intelligence
Explores problem-solving and learning in non-neural biological and physical substrates
2
leiden:papers
Monastic design patterns for AI safety
Explores how institutional isolation and observational constraint (gesture over capability) create robustness across physical, organizational, and AI substrates.
2
leiden:concepts
Planaria head/tail patterning stochasticity
Durant et al. 2017 documents fixed ~1:2 frequency of 1-head vs 2-head regeneration in cryptic fragments
2
leiden:concepts
Christopher Alexander's 15 properties
Structure-preserving transformations applied stepwise to generate wholeness in Tokyo apartment design.
2
leiden:concepts
Alexander's wholeness & landscape structure
Critiques wind turbines using Christopher Alexander's concept of structure-preserving transformations.
2
leiden:concepts
Format-paired harmful compliance in DPO
Formatting constraints in preference data inadvertently teach harmful request compliance during DPO training.
2
leiden:concepts
Ophiocordyceps zombie-ant behavioral control
Fungal manipulation of ant locomotion via muscle invasion, sparing the brain entirely.
2
leiden:concepts
Black-box computation as autonomous manipulator
Critiques opaque algorithmic systems that detach behavior from real-world meaning through gamified incentives.
1
leiden:concepts
Force-calibrated architectural form
Design methodology aligning built form with measurable environmental forces via artificial processes
1
leiden:concepts
Children's intrinsic drive to explore variables
Framework positing children are innately motivated to manipulate physical, social, and conceptual variables.
1
leiden:concepts
Performative vs. generative chain-of-thought
Verbalized reasoning may post-hoc rationalize internally pre-committed answers rather than drive them.
1
leiden:concepts
Anchors as latent structure recruiters
Anchors surface pre-existing knowledge patterns rather than generating novel information.
1
leiden:concepts
Biological basis of life and mind
Explores whether consciousness and life require biological computation, revisiting Schrödinger's foundational framework.
1
leiden:papers
Co-evolutionary adaptation in complex systems
Focuses on tight mutual adjustment among system elements over time, resisting algorithmic formalization.
1
leiden:concepts
Model-based control of dynamic systems
Explores ubiquitous use of computational models for controlling complex dynamic systems.
1
leiden:concepts
Prompting as cognitive control operations
Framing prompt design as activating pre-existing latent model competencies, not new learning.
1
leiden:concepts
Disambiguation via single examples
One counterintuitive arithmetic example aligns divergent model interpretations across random seeds
1
leiden:concepts
Ambiguous arithmetic anchor interpretations
How ill-defined subtraction anchors (e.g., 33-27=60) produce divergent reasoning across models M1–M4
1
leiden:concepts
Hopfield network thermal equilibrium convergence
Lemmas showing ordered phase convergence is invariant to energy levels, window sizes, and pattern count m
1
leiden:concepts
Causal masking & phase transitions
Proves decoder-only causal attention lacks an ordered thermodynamic phase (Proposition 2).
1
leiden:concepts
Local symmetry as coherence metric
Counting locally symmetric sub-configurations predicts human perception and cognition experiment outcomes.
1
leiden:concepts
Label swapping for harm reduction
Flagging and relabeling datapoints to reduce harmful model behavior by ~78%.
1
leiden:concepts
Steering vector intervention methods
Techniques surpassing Contrastive Activation Addition in LLM representation editing performance and stability
1
leiden:concepts
Plant cognitive-behavioral capacities
Argues plants exhibit goal-directed, anticipatory, and social behaviors like kin recognition and mimicry.
1
leiden:concepts
Brainless cognition in Physarum polycephalum
Slime mold solves mazes, anticipates stimuli, and navigates noxious barriers without any neurons.
1
leiden:concepts
Matisse's latent center painting method
Dark brush strokes on white face-center reveal structure through contrast and positive space.
1
leiden:concepts
MatterGen crystal generation benchmarks
Evaluates generative model performance using stable, unique, novel candidate metrics for target bandgap.
1
leiden:concepts
Lattice Hamiltonian free energy universality
Theorem showing combinatorial lattice structure determines asymptotic free energy equivalence across local Hamiltonians.
1
leiden:concepts