{"paper":{"title":"AdapShot: Adaptive Many-Shot In-Context Learning with Semantic-Aware KV Cache Reuse","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"AdapShot selects the optimal number of in-context examples for each query by measuring output entropy in a probe run and reuses KV cache with reordering to enable efficient many-shot learning.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Jie Ou, Jinyu Guo, Ruiqi Wu, Shiyao Guo, Wenhong Tian, Wenyi Li, Yuang Li, Zhaokun Wang","submitted_at":"2026-05-05T11:16:52Z","abstract_excerpt":"Many-Shot In-Context Learning (ICL) has emerged as a promising paradigm, leveraging extensive examples to unlock the reasoning potential of Large Language Models (LLMs). However, existing methods typically rely on a predetermined, fixed number of shots. This static approach often fails to adapt to the varying difficulty of different queries, leading to either insufficient context or interference from noise. Furthermore, the prohibitive computational and memory costs of long contexts severely limit Many-Shot's feasibility. To address the above limitations, we propose AdapShot, which dynamically"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Extensive experiments demonstrate that AdapShot achieves an average performance gain of around 10% and a 4.64x speedup compared to state-of-the-art DBSA.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That output entropy from a short probe run is a sufficient and unbiased signal for selecting the globally optimal shot count, and that the decoupling-plus-re-encoding step for KV cache reordering introduces no accuracy degradation.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"AdapShot adaptively tunes shot count via entropy probes and reuses semantically-matched KV caches with position decoupling to deliver ~10% accuracy gains and 4.64x speedup over fixed-shot baselines.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"AdapShot selects the optimal number of in-context examples for each query by measuring output entropy in a probe run and reuses KV cache with reordering to enable efficient many-shot learning.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f07e8d7cb025c50c88c0e2b6354bb4dfa0766f474c031be44478b97bdf8624ba"},"source":{"id":"2605.03644","kind":"arxiv","version":2},"verdict":{"id":"16c1b36e-cfe9-4013-abb2-c6f18cbb29ab","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-07T16:34:19.320364Z","strongest_claim":"Extensive experiments demonstrate that AdapShot achieves an average performance gain of around 10% and a 4.64x speedup compared to state-of-the-art DBSA.","one_line_summary":"AdapShot adaptively tunes shot count via entropy probes and reuses semantically-matched KV caches with position decoupling to deliver ~10% accuracy gains and 4.64x speedup over fixed-shot baselines.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That output entropy from a short probe run is a sufficient and unbiased signal for selecting the globally optimal shot count, and that the decoupling-plus-re-encoding step for KV cache reordering introduces no accuracy degradation.","pith_extraction_headline":"AdapShot selects the optimal number of in-context examples for each query by measuring output entropy in a probe run and reuses KV cache with reordering to enable efficient many-shot learning."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.03644/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T13:36:56.140453Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-20T00:31:21.764638Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T15:07:50.971941Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"ac8b52b7d1ef58bd56ee4d3b83590fefb6bd1fb7a971516da15d40e3a74a011f"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}