concept
active
concept:temporal-straighteningTemporal straightening
Linearization of representational trajectories observed in brains and LMs.
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Interpretation of smooth predictive trajectories.
Methods (1)
method
- Splits activations at critical frequency into equal-energy halves.
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.
- Control procedure that permutes the concatenation order of (C)ARR while preserving internal token order; repeated 10 times.
- In active inference, distant future predictions are naturally less precise, leading to Bayes-optimal discounting without an explicit discount factor.
- Mechanistic interpretability technique for locating factual associations, mentioned as future work direction.
- Method by Chan et al. 2022 for rigorously testing interpretability hypotheses via interventions
- Proposed objective decomposing activations into predictable and novel components.
- The adaptive, incremental nature of living process, allowing small steps with continuous evaluation and adjustment.
- Confound where naming injected concepts reflects direct logit effects rather than metacognitive awareness, raised by Morris & Plunkett