{"paper":{"title":"Dynamic Mixed-Precision Routing for Efficient Multi-step LLM Interaction","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Dynamic mixed-precision routing lets LLMs switch between high and low precision at each step to cut costs while keeping task success high.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Huanrui Yang, Jianing Deng, Jingtong Hu, Song Wang, Tianlong Chen, Yuanzhe Li","submitted_at":"2026-02-02T19:24:04Z","abstract_excerpt":"Large language models (LLMs) achieve strong performance in long-horizon decision-making tasks through multi-step interaction and reasoning at test time. While practitioners commonly believe a higher task success rate necessitates the use of a larger and stronger LLM model, multi-step interaction with a large LLM incurs prohibitive inference cost.\n  To address this problem, we explore the use of low-precision quantized LLMs in the long-horizon decision-making process. Based on the observation of diverse sensitivities among interaction steps, we propose Dynamic Mixed-Precision Routing (DMR), a f"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments on ALFWorld and WebShop demonstrate that our approach achieves a strong accuracy-cost trade-off over single-precision baselines.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The observation that interaction steps have diverse sensitivities to precision allows a router to be trained that reliably selects the right precision level without harming overall task success.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"DMR uses a router trained in two stages (KL-divergence supervision then GRPO) to pick high or low precision LLMs per step, delivering better accuracy-cost trade-offs than fixed-precision baselines on ALFWorld and WebShop.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Dynamic mixed-precision routing lets LLMs switch between high and low precision at each step to cut costs while keeping task success high.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"0cab96d6ac0b07af14e1e5a4908ba364d2a65dbde47c058901cbabd74cf20275"},"source":{"id":"2602.02711","kind":"arxiv","version":2},"verdict":{"id":"e67e2e07-e25d-4b79-8827-dadc1b5f6d3d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T07:52:44.077711Z","strongest_claim":"Experiments on ALFWorld and WebShop demonstrate that our approach achieves a strong accuracy-cost trade-off over single-precision baselines.","one_line_summary":"DMR uses a router trained in two stages (KL-divergence supervision then GRPO) to pick high or low precision LLMs per step, delivering better accuracy-cost trade-offs than fixed-precision baselines on ALFWorld and WebShop.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The observation that interaction steps have diverse sensitivities to precision allows a router to be trained that reliably selects the right precision level without harming overall task success.","pith_extraction_headline":"Dynamic mixed-precision routing lets LLMs switch between high and low precision at each step to cut costs while keeping task success high."},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"72ddd5294c4e2f5fc1657334bc660c5bca43a59e53677028fc119b6f052bad21"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}