{"slug": "stop-when-further-reasoning-won-t-help-attention-state-adaptive-generation-in", "title": "Stop When Further Reasoning Won't Help: Attention-State Adaptive Generation in Reasoning Models", "summary": "Researchers propose ASAG, a training-free method that monitors attention distributions to detect when a reasoning model has reached a conclusion, stopping generation early. Applied to DeepSeek-R1-Distill and Qwen3 models, ASAG improves average accuracy by 3.2% while reducing generated tokens by nearly 40% on Qwen3-8B across nine benchmarks.", "body_md": "arXiv:2606.15070v1 Announce Type: new\nAbstract: By incorporating test-time compute scaling, large reasoning models (LRMs) can solve complex problems through explicit chain-of-thought (CoT) reasoning processes. However, they often suffer from overthinking, resulting in redundant token outputs and degraded accuracy. Current methods to mitigate this issue remain limited: training-based approaches require substantial computational resources, while training-free methods rely on well-crafted prompts or unreliable confidence signals. In this work, we investigate early stopping from the perspective of attention distributions and propose a simple method, ASAG, which infers the model's reasoning state and adaptively adjusts the generation strategy. The proposed framework is training-free and plug-and-play, enabling seamless integration into existing LRMs. Extensive experiments on nine benchmarks demonstrate consistent improvements across mainstream LRMs with varying parameter scales, including the DeepSeek-R1-Distill and Qwen3 series. Specifically, ASAG improves average accuracy by 3.2% while reducing the number of generated tokens by nearly 40% across all reasoning tasks on Qwen3-8B.", "url": "https://wpnews.pro/news/stop-when-further-reasoning-won-t-help-attention-state-adaptive-generation-in", "canonical_source": "https://arxiv.org/abs/2606.15070", "published_at": "2026-06-16 04:00:00+00:00", "updated_at": "2026-06-16 04:23:22.565928+00:00", "lang": "en", "topics": ["large-language-models", "ai-research", "natural-language-processing"], "entities": ["DeepSeek-R1-Distill", "Qwen3", "Qwen3-8B", "ASAG"], "alternates": {"html": "https://wpnews.pro/news/stop-when-further-reasoning-won-t-help-attention-state-adaptive-generation-in", "markdown": "https://wpnews.pro/news/stop-when-further-reasoning-won-t-help-attention-state-adaptive-generation-in.md", "text": "https://wpnews.pro/news/stop-when-further-reasoning-won-t-help-attention-state-adaptive-generation-in.txt", "jsonld": "https://wpnews.pro/news/stop-when-further-reasoning-won-t-help-attention-state-adaptive-generation-in.jsonld"}}