{"slug": "kv-prm-efficient-process-reward-modeling-via-kv-cache-transfer-for-multi-agent", "title": "KV-PRM: Efficient Process Reward Modeling via KV-Cache Transfer for Multi-Agent Test-Time Scaling", "summary": "Researchers introduced KV-PRM, a process reward model that reduces scoring cost from O(L²) to O(L) by leveraging KV cache from LLM generation, achieving up to 5,000x reduction in FLOPs and 37x reduction in latency while matching or outperforming text-based PRMs on MATH, GSM8K, and AIME benchmarks.", "body_md": "arXiv:2607.09153v1 Announce Type: new\nAbstract: Process Reward Models (PRMs) have been proven to be highly effective in guiding test-time scaling (TTS) methods, which significantly boost the capabilities of LLM-based multi-agent systems. However, existing PRMs are text-based: they re-encode the entire trajectory text from scratch. In long multi-agent rollouts, the scoring cost, growing quadratically with respect to sequence length L, creates a severe computational bottleneck, severely limiting PRMs' application in long-context scenarios. To resolve this, we introduce KV-PRM, a highly efficient process reward model that eliminates the heavy text re-encoding by directly reading the KV cache produced naturally during the LLM's generation phase. By processing a single \"verify token\" against the pre-existing KV cache, KV-PRM reduces the scoring cost from O(L^2) to O(L). We formally prove that the KV cache contains strictly greater information capacity than text, and is more efficient for downstream reward modeling. Empirically, across the MATH, GSM8K, and AIME benchmarks, KV-PRM matches or strictly outperforms text-PRMs under various TTS methods such as Beam Search, MCTS, and Weighted Voting, with up to a 5,000x reduction in scoring FLOPs, a 37x reduction in latency, and a 34x reduction in per-sequence memory footprint compared to text-based PRMs.", "url": "https://wpnews.pro/news/kv-prm-efficient-process-reward-modeling-via-kv-cache-transfer-for-multi-agent", "canonical_source": "https://arxiv.org/abs/2607.09153", "published_at": "2026-07-13 04:00:00+00:00", "updated_at": "2026-07-13 04:08:27.365130+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-research", "ai-infrastructure"], "entities": ["KV-PRM", "MATH", "GSM8K", "AIME"], "alternates": {"html": "https://wpnews.pro/news/kv-prm-efficient-process-reward-modeling-via-kv-cache-transfer-for-multi-agent", "markdown": "https://wpnews.pro/news/kv-prm-efficient-process-reward-modeling-via-kv-cache-transfer-for-multi-agent.md", "text": "https://wpnews.pro/news/kv-prm-efficient-process-reward-modeling-via-kv-cache-transfer-for-multi-agent.txt", "jsonld": "https://wpnews.pro/news/kv-prm-efficient-process-reward-modeling-via-kv-cache-transfer-for-multi-agent.jsonld"}}