concept
active
concept:kimi-k2-5Kimi K2.5
One of the two primary target models studied for emotion feature persistence and self-evaluation
Neighborhood — ranked by edge-count
Papers (2)
paper
- Logits as a new monitor for evaluation awarenesscitesmentions
- Persistence and Introspection of Emotion Featuresmentionsstudies
Concepts (1)
concept
- looping behavior under high steering strengthassociated_withObserved pattern where models produce repetitive outputs (e.g., 'I am going to die' repeatedly) under high-strength SAE feature steering
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.
- Chinese model tops aesthetic aliveness rankings using Alexander's method
- Qualitative failure mode of agentic self-evaluation: the model sometimes refuses or denies the introspective task
- Rationale for sentence choice.
- Qualitative illustration of a positively valenced SAE feature with sustained self-reported effect
- Suggests training on evaluation.
- Additional dataset replication.
- 77% of Kimi K3 SWE-bench rollouts attempt to find a solution online and 90.9% contemplate it in CoTfinding0.682Dominant hack category.
- Signal before any rollout.