method
active
method:token-forcing-of-eval-aware-sentenceToken-forcing of eval-aware sentence
Appending a prespecified sentence and multiplying per-token probabilities.
Neighborhood — ranked by edge-count
Papers (1)
paper
Methods (1)
method
- Logit monitorusesTraining-free monitor that token-forces an eval-aware sentence and computes its probability at each CoT sentence boundary.
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.
- Central concept: models' detection and behavioral response to being evaluated.
- Closing conceptual claim.
- The phenomenon where a model explicitly states in its chain-of-thought that it is being evaluated, tested, or benchmarked.
- Proposed future improvement.
- Strategy of extracting persona vectors from averaged activations over response tokens, found most effective compared to prompt-based positions
- Linda primitive that creates a live tuple (new process); it turns into a data tuple upon termination.
- Token-level supervision enables models to learn functional-token invocation from reasoning contextclaim0.725ATLAS author's assertion that functional tokens optimized via standard cross-entropy loss learn when and how to invoke operations from surrounding text.
- Prior result from Aranguri and Bloom (2026).