arXiv:2606. 04160v1 Announce Type: cross Abstract: Safety alignment in instruction-tuned large language models (LLMs) depends on a model's ability to reliably refuse to respond to harmful or disallowed requests.
arXiv:2606.04160v2 Announce Type: replace-cross Abstract: Safety alignment in instruction-tuned large language models (LLMs) depends on a model's ability to reliably refuse harmful or disallowed requ...
arXiv:2603.13359v2 Announce Type: replace Abstract: Language models are commonly fine-tuned for safety alignment to refuse harmful prompts. One approach fine-tunes them to generate categorical refusa...
arXiv:2608.30197v1 Announce Type: new Abstract: Safety alignment is essential for deploying large language models, requiring systems to prevent harmful compliance while preserving helpfulness on beni...
The paper addresses the problem of over‑refusal in safety‑aligned large language models, where benign instructions are incorrectly rejected. It identifies that a small set of hypersensitive safety heads in transformer attention misfire on hard‑safe prompts, causing abnormal attention entanglement and high‑entropy routing conflicts that block necessary attention to target entities. To mitigate this, the authors propose Semantic Routing Calibration (SRC), a lightweight, training‑free inference framework that dynamically suppresses these hypersensitive heads and fuses logits from dual branches to restore trustworthy reasoning while preserving intrinsic safety performance.
The paper introduces Weighted Refusal Direction Optimization (WRDO), a method that constructs a composite refusal direction by learning aggregation weights over per‑prompt refusal directions. WRDO aims to strengthen refusal interventions in aligned language models by allowing heterogeneous harmful prompts to contribute unequally. Experiments on three open‑weight models, using guard judges and HarmBench, demonstrate that weighted aggregation outperforms single‑direction methods and can also be used in reverse to restore refusal behavior in a damaged checkpoint.
The paper addresses the problem of large language models (LLMs) over-refusing to comply with benign safety-related instructions. It identifies that a small number of hypersensitive safety heads in transformer attention misfire on hard-safe prompts, causing abnormal attention entanglement and high-entropy routing conflicts that lead to refusals. The authors propose Semantic Routing Calibration (SRC), a lightweight, training‑free inference framework that dynamically suppresses these problematic heads and fuses logits to restore trustworthy reasoning while preserving intrinsic safety performance.
arXiv:2609.00760v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly trained to decline queries that fall outside their knowledge (knowledge-based refusal, KR) or violate saf...
The paper introduces RARE, a router‑agnostic representation engineering framework for Mixture‑of‑Experts language models. RARE projects behavioral perturbations onto the null space of the router matrix to avoid affecting routing, and corrects downstream routing drift. Experiments on six open‑weight MoE models show that RARE improves steering tasks—reducing harmfulness, increasing truthfulness, and enhancing factual editing—while preserving overall model accuracy.
arXiv:2610.02910v1 Announce Type: new Abstract: Suppressing a small set of routed experts can weaken the safety behavior of a sparse Mixture-of-Experts (MoE) language model without retraining. Which...
The paper investigates whether large language models’ refusal behavior in text transfers to tool‑calling contexts. A benchmark of 2,304 prompts across four domains compares text‑only, standard tool‑enabled, and adversarially framed tool‑enabled settings, revealing that 10–52 % of harmful requests refused in chat still lead to unsafe tool calls. The authors identify a refusal‑associated direction in the models’ residual stream, showing that lower activation along this direction predicts unsafe calls, but restoring the signal only partially mitigates the issue and often suppresses benign tool use.
The paper introduces RASET, a router‑agnostic safety‑critical expert tuning framework for Mixture‑of‑Experts (MoE) large language models. RASET identifies a small subset of experts that are responsible for safety enforcement and applies parameter‑efficient tuning only to those experts, preserving the model’s intrinsic routing behavior. Experiments on five open‑weight MoE backbones show that RASET achieves a high safety‑bypass yield, outperforming existing baselines by a significant margin.
The paper investigates how large language models balance helpfulness and safety by refusing harmful queries while responding to benign ones. It decomposes safety-tuning responses into a boilerplate refusal statement and a rationale, finding that the statement causes false refusals by relying on superficial cues. Training on rationales alone reduces false refusals without compromising safety performance, suggesting that fine‑grained safety supervision is essential for better alignment.