claim
active
claim:behavioral-signatures-reflect-beliefs-held-while-reading-since-behavior-is-predictable-from-activations-without-concept-instructionBehavioral signatures reflect beliefs held while reading, since behavior is predictable from activations without concept instruction
Validates behavioral probe.
Source paper
extracted_from(2026) · Eric Bigelow · Raphaël Sarfati · Daniel Wurgaft · Owen Lewis +4
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (1)
finding
- Beliefs linearly decodable.
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.
- Future work.
- Extension of the thesis to deployed LLM inference via in-context learning
- The central objection the paper must answer to establish identity over mere correlation
- Authors' caveat that conversational context persistence rather than internal emotion state persistence could explain findings
- Acknowledged alternative explanation that the paper does not rule out
- Caveat on interpretation.
- Key prescriptive statement supporting the system-agnostic approach.