anchor
Findings
- IC-1015GPT-J and 10 other LLMs exhibit overthinking: calibrated accuracy given incorrect few-shot demonstrations peaks at a critical layer then declines, and ablating 5 false induction heads in late layers reduces the accuracy gap by 38.9% on average [eval]
- IC-1088ICL predictions in LLaMA, LLaMA-2, and Falcon models depend on in-context label information and can learn truly novel label relationships [eval]
- IC-1089ICL in LLaMA, LLaMA-2, and Falcon models cannot fully overcome pre-training label preferences when in-context labels are flipped [eval]
- IC-1090ICL in LLaMA, LLaMA-2, and Falcon models preferentially uses in-context label information closer to the query rather than treating all examples equally [eval]
- IC-1206GPT-2-XL, GPT-J, Falcon-7B, Llama-2-7B, and Llama-2-13B are vulnerable to backdoor injection via lightweight parameter editing with only 15 samples, achieving near-100% attack success rate while preserving clean performance [eval]
- IC-1208Larger LLMs (Llama-2-13B) require more data samples for successful backdoor injection via parameter editing compared to smaller models (GPT-2-XL 1.5B) [eval]
- IC-327Llama-3-8B-Instruct, Mistral-7B-Instruct-v0.3, and several other LLMs produce well-calibrated verbal confidence estimates on classification tasks [eval]
- IC-328Llama-3-8B-Instruct and Mistral-7B-Instruct-v0.3 are susceptible to confidence-elicitation-guided word substitution attacks, with CEAttack outperforming existing hard-label black-box methods [eval]
- IC-329GPT-4o is more robust to confidence-elicitation-guided word substitution attacks than open-source LLMs, with lower attack success rates and better confidence calibration [eval]
- IC-578LLMs encode input text as linearly separable representations in forerunner token hidden states, emerging in early layers and enhanced by in-context demonstrations [eval]
- IC-580The 3-step ICL inference circuit (text encoding, semantics merge, feature retrieval) is a dominant causal mechanism, as ablating the corresponding attention connections significantly degrades ICL accuracy [eval]
- IC-581Induction heads for ICL operate on task-specific attention subspaces, with partial overlap across tasks, and the geometry of these subspaces explains demonstration saturation [eval]