institute
active
concept:anthropicAnthropic
Lab behind Claude models and Constitutional AI training approach; represents highest baseline scores and lowest prompt lift.
Neighborhood — ranked by edge-count
Thinkers (56)
thinker
- Chris Olahaffiliated_with
- Jack Lindseyaffiliated_with
- Neel Nandaaffiliated_with
- Ethan Perezaffiliated_with
- Trenton Brickenaffiliated_with
- Samuel Marksaffiliated_with
- Evan Hubingeraffiliated_with
- Joshua Batsonaffiliated_with
- Jared Kaplanaffiliated_with
- Adam Karvonenaffiliated_with
- Wes Gurneeaffiliated_with
- Kyle Fishaffiliated_with
- Nelson Elhageaffiliated_with
- Subhash Kantamneniaffiliated_with
- Christopher Olahaffiliated_with
- Emmanuel Ameisenaffiliated_withCo-author; enabled feature-level attributions, multi-step inference analysis.
- Nicholas Sofroniewaffiliated_with
- Carson Denisonaffiliated_with
- Isaac Kauvaraffiliated_with
- Adam Pearceaffiliated_with
- Euan Ongaffiliated_with
- Tom Henighanaffiliated_with
- Adam Jermynaffiliated_with
- Catherine Olssonaffiliated_with
+32 more
Papers (28)
paper
- Open Problems in Mechanistic Interpretabilitycitesmentions
- Toy Models of Superpositioncitesmentions
- On the Biology of a Large Language Modelcitesmentions
+4 more
Communities (1)
community
- LLM Introspectionmembers_of
Frameworks (1)
framework
- Constitutional AIassociated_withAlignment approach by Anthropic that explicitly trains self-observation; predicts highest baseline and lowest prompt lift.
Datasets (3)
dataset
- Claude Opus 4.6associated_withPrimary target model for NLA development and case studies; underwent pre-deployment audit using NLAs.
- Claude Haiku 3.5associated_withTarget model for NLA training and evaluation; used in quantitative benchmarking.
- Claude Haiku 4.5associated_withTarget model for NLA training and evaluation; shows similar FVE curves to Haiku 3.5.
Artifacts (2)
artifact
- Claude 4.1 Opusassociated_withLLM judge and explainer used throughout.
- Anthropic open-source visualization tools used for graphs and max-activating examples.
Concepts (1)
concept
- Emotion Concepts and their Function in a Large Language ModelauthoredmentionsThe prior Anthropic paper whose findings about emotion features in Claude this paper builds upon and extends
Methods (1)
method
- Circuit tracingstudiesAnthropic technique revealing computational pathways in Claude.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- The belief that human intelligence is the norm, against which other intelligences are measured.
- The team that produced this developing-work update.
- A generic concept of intelligence not modeled on human cognition, encompassing all life and potential machines.
- Tendency to under-attribute human traits to nonhumans.
- Tendency to over-attribute human traits to nonhumans.
- Tendency of the model to recruit human-like mental concepts when representing its assistant persona.
- The fact that contemporary AI is trained to reproduce human behavioural/cognitive outputs, confounding behavioural indicators.
- Authors' affiliation.