concept
active
concept:backtrackingBacktracking
Reasoning behavior for which the SAEs learned features.
Neighborhood — ranked by edge-count
Papers (1)
paper
Concepts (1)
concept
- Backtracking Latentsrelated_toSAE latents that rise as correction approaches and peak after self-correction begins, complementing OTDs
Findings (1)
finding
- R1 general reasoning SAE learned features reconstructing reasoning behaviors such as backtrackingcitesCurated features show core reasoning behaviors are captured.
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.
- A sequence where no step forces the undoing of previous steps; these sequences lie at the core of the theory of living process and can be identified experimentally.
- Claims that although a purely mathematical identification method is lacking, a well-defined experimental procedure exists to find good sequences.
- Supports the awareness hypothesis.
- Standard learning algorithm for deep neural networks that propagates error signals to adjust weights; lacks convergence guarantee for non-linearly separable functions
- The training method of modern AI systems; each step computes goal-relative error identified with valence
- Interpretive claim linking observable CoT behaviors to genuine internal uncertainty shifts
- The process of reinterpreting and recontextualizing memory engrams for new bodies and environments.
- Model selects future output targets and works backward from them.