{"slug": "polestar-drift-aware-cache-calibration-and-token-commitment-for-efficient-of", "title": "Polestar: Drift-Aware Cache Calibration and Token Commitment for Efficient Inference of Diffusion LLMs", "summary": "Researchers introduced Polestar, a training-free inference framework for diffusion large language models (dLLMs) that uses token representation drift to improve efficiency. Polestar comprises two components—Polestar-Cache for sparse KV-cache refreshes and Polestar-Commit for identifying commit-ready tokens—achieving up to 10.73% accuracy improvement and 3.7x higher throughput on mathematics and coding benchmarks.", "body_md": "arXiv:2607.14107v1 Announce Type: new\nAbstract: The inference efficiency of diffusion large language models (dLLMs) is constrained by two challenges: bidirectional attention precludes efficient KV-cache reuse, while increasing decoding parallelism with static confidence thresholds can compromise generation quality. We observe that both challenges arise from a shared phenomenon: as tokens are decoded, their contextual integration through bidirectional attention causes token representations to drift (evolve) across decoding steps. This insight motivates Polestar, a training-free inference framework that uses token representation drift as a unified signal to jointly address both challenges. Polestar comprises two components: Polestar-Cache, which identifies stale KV-cache positions via drift and performs sparse KV-cache refreshes to enable efficient reuse, and Polestar-Commit, which detects sharp drift events to reliably identify commit-ready tokens. Across mathematics and coding benchmarks on several dLLM families, Polestar sets a new state of the art on the accuracy-throughput Pareto frontier, achieving up to 10.73% accuracy improvement, up to 3.7x higher throughput, and high decoding parallelism of 3.67 tokens per forward pass over existing baselines.", "url": "https://wpnews.pro/news/polestar-drift-aware-cache-calibration-and-token-commitment-for-efficient-of", "canonical_source": "https://arxiv.org/abs/2607.14107", "published_at": "2026-07-17 04:00:00+00:00", "updated_at": "2026-07-17 04:04:17.230966+00:00", "lang": "en", "topics": ["large-language-models", "artificial-intelligence", "machine-learning", "ai-infrastructure"], "entities": ["Polestar", "Polestar-Cache", "Polestar-Commit"], "alternates": {"html": "https://wpnews.pro/news/polestar-drift-aware-cache-calibration-and-token-commitment-for-efficient-of", "markdown": "https://wpnews.pro/news/polestar-drift-aware-cache-calibration-and-token-commitment-for-efficient-of.md", "text": "https://wpnews.pro/news/polestar-drift-aware-cache-calibration-and-token-commitment-for-efficient-of.txt", "jsonld": "https://wpnews.pro/news/polestar-drift-aware-cache-calibration-and-token-commitment-for-efficient-of.jsonld"}}