# Semantic Routing Calibration mitigates LLM over-refusal

Digest AI · Research · published 2026-09-23T04:00:00Z

Canonical: https://digestai.news/story/semantic-routing-calibration-mitigates-llm-over-refusal

## Summary

Large language models that are aligned for safety often reject benign instructions, a problem known as over‑refusal. Prior work has linked the issue to static representation overlap, but the underlying dynamic causes have not been fully explored. The paper investigates over‑refusal by examining internal routing conflicts in transformer attention, focusing on a small set of hypersensitive safety heads that misfire on hard‑safe prompts.

To address this, the authors introduce Semantic Routing Calibration (SRC), a lightweight, training‑free inference framework. SRC identifies the problematic safety heads at inference time and dynamically suppresses them. It also applies a dual‑branch logits fusion that acts as a safety regularizer during decoding, helping the model maintain trustworthy reasoning.

Extensive experiments show that SRC reduces over‑refusal rates while keeping intrinsic safety performance largely intact. The authors claim that the approach preserves safety constraints and improves the model’s ability to handle benign safety‑related queries.

## Key points

- SRC suppresses hypersensitive safety heads during inference.
- SRC uses dual‑branch logits fusion as a safety regularizer.
- Experiments show SRC reduces over‑refusal while preserving safety performance.

## Why it matters

Reducing over‑refusal improves user experience and trust in safety‑aligned LLMs, enabling more reliable responses to benign safety‑related queries without compromising safety constraints.

## Sources

1. [Mitigating LLM Over-Refusal via Dynamic Semantic Routing Calibratione](https://arxiv.org/abs/2609.25049) (arXiv cs.CL, 2026-09-23, primary source)

## Cite

Digest AI, "Semantic Routing Calibration mitigates LLM over-refusal", 23 September 2026, https://digestai.news/story/semantic-routing-calibration-mitigates-llm-over-refusal

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/semantic-routing-calibration-mitigates-llm-over-refusal.json