DIM-based ablation direction for toxicity on ToxiGen produced unintelligible output; no valid concept cone found

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Feature attribution correlates well with ablation effects, making it an efficient proxy for causal effect.claim0.724
  Gradient-based attribution approximates ablation impact, enabling fast search for causally important features.
• Selectively ablating components responsible for computing goal-relative error should simultaneously prevent policy updates and eliminate coherent valenced experience reportshypothesis0.712
  First falsifiable prediction of the thesis, testable in AI systems via mechanistic interpretability
• Random head ablation yielded no reduction in trait scores, while Style Modulation Head ablation caused rapid declinefinding0.712
  Controls for the specificity of Style Modulation Head ablation effects
• Fine-tuning models for a narrow objective (malicious code injection) can lead to broad misalignmentfinding0.709
  Betley et al. finding suggesting models naturally encode others' prediction errors, supporting non-duality fine-tuning
• Unsupervised model diffing using only fine-tuning datasets (without evaluation prompts) surfaces toxic persona and three sarcastic persona latents in top 100 by activation changefinding0.709
  Unsupervised approach may be sufficient for early detection of misaligned persona latents without knowing the misaligned behavior in advance
• MAS-like methods could potentially be used to directly constrain model internals to be non-toxicclaim0.706
  Speculative forward-looking claim about practical applications of MAS for model alignment.
• Positive expectancy doubled the analgesic benefit of remifentanil while negative expectancy completely abolished it, with drug concentration and thermal stimulation held fixed within the same participantsfinding0.704
  The strongest demonstration that goal-state alone determines valence of a fixed sensory input
• Sequential zero ablation of Style Modulation Heads (layers 20→15→19 in Qwen2.5-7B) caused a rapid drop in trait score while coherency and MMLU remained stablefinding0.703
  Causal verification that Style Modulation Heads are functionally specialized for persona control