hallucination

Model tendency to generate incorrect intermediate reasoning steps that mislead answer inference, particularly in 1B-models.

Neighborhood — ranked by edge-count

paper

claim

Vision features enable generation of more effective rationales that reduce hallucination and improve answer inference
associated_with
Core interpretive assertion: multimodal information (vision + language) produces higher-quality intermediate reasoning steps compared to language-only approaches.

concept

Strategic Deception
contradicts
Central concept of the paper: deliberate, goal-driven deception where model reasoning contradicts outputs

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

controlled hallucinationconcept0.822
Anil Seth's term for perception as brain's best guess, filling gaps.
Hallucination in LLMsconcept0.764
Problem cited as a shortcoming of current LLMs; PRH predicts hallucinations should decrease with scale
Perceptionconcept0.737
Equated with inference of past, present and future hidden states via minimization of variational free energy.
Consciousnessconcept0.729
Core concept: capacity to experience as a subject; argued to be substrate-independent and achievable across diverse biological systems.
Consciousness (phenomenal consciousness)concept0.715
The state of having subjective experiences; there is something it is like to be the subject.
Awakeningconcept0.714
Buddhist concept formalised as embodied recognition that no finite agent can evidence its separability; equated with pruning sigma via BMR
shatteringconcept0.712
The phenomenon where SAEs break a smooth geometric manifold into many small, seemingly unrelated pieces, losing overarching structure.
Perception-Action Cycleconcept0.710
Circular causality between perception and action; central to enactive interpretation