method
active
method:self-supervised-context-predictionSelf-supervised context prediction
Predict preceding or following text from an activation; scalable without labeled data.
Neighborhood — ranked by edge-count
Papers (1)
paper
Frameworks (1)
framework
- Framework training LLMs to answer questions about externally-provided activation vectors
Claims (1)
claim
- Justifies scalability of the task.
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 thesis that transformers develop a self-model via ICL, not only from training data; base models bootstrap self-referential reasoning.
- Reports phase-like breakpoints and geometry changes as context scales; UCCT provides measurable predictor
- Feature that activates across all tokens within a specific context (e.g., DNA sequences, base64 strings)
- Central interpretive claim from statistical analysis
- NLI class indicating one sentence logically follows from another; weighted -1 in NLI Diversity scoring
- The key novel contribution: an internal model that learns the density of familiar multisensory experiences and drives mark-removal behavior through mismatch with the free energy principle
- NLI class indicating two sentences are semantically incompatible; weighted +1 in NLI Diversity scoring
- The ability of a model to observe its own state, measured by Koan Battery; can be lifted by contemplative prompts.