{"paper":{"title":"Precise Debugging Benchmark: Is Your Model Debugging or Regenerating?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Frontier LLMs pass debugging tests above 76 percent yet edit with precision below 45 percent, often regenerating entire solutions instead of making targeted fixes.","cross_cats":["cs.CL"],"primary_cat":"cs.SE","authors_text":"Honghua Dong, Miaosen Chai, Robin Jia, Shangshang Wang, Song Bian, Wang Bill Zhu, Willie Neiswanger, Yejia Liu","submitted_at":"2026-04-19T09:08:23Z","abstract_excerpt":"Unlike code completion, debugging requires localizing faults and applying targeted edits. We observe that frontier LLMs often regenerate correct but over-edited solutions during debugging. To evaluate how far LLMs are from precise debugging, we introduce the Precise Debugging Benchmark (PDB) framework, which automatically converts any coding dataset into a debugging benchmark with precision-aware evaluation. PDB generates buggy programs by synthesizing verified atomic bugs and composing them into multi-bug programs. We define two novel metrics, edit-level precision and bug-level recall, which "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"frontier models, such as GPT-5.1-Codex and DeepSeek-V3.2-Thinking, achieve unit-test pass rates above 76% but exhibit precision below 45%, even when explicitly instructed to perform minimal debugging.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The automatically synthesized verified atomic bugs and their compositions accurately represent real-world debugging scenarios, and the new edit-level precision and bug-level recall metrics validly measure 'precise debugging' behavior.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Frontier LLMs pass unit tests over 76% of the time on debugging tasks but achieve edit precision below 45%, indicating regeneration rather than precise debugging.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Frontier LLMs pass debugging tests above 76 percent yet edit with precision below 45 percent, often regenerating entire solutions instead of making targeted fixes.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a0d6bb01a10e7ad1e8db6572a092252383a3053a08787a1d072936f35a79b5c4"},"source":{"id":"2604.17338","kind":"arxiv","version":4},"verdict":{"id":"e674bdeb-e6d4-473d-a1ed-67e9a97eae90","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T06:02:06.759620Z","strongest_claim":"frontier models, such as GPT-5.1-Codex and DeepSeek-V3.2-Thinking, achieve unit-test pass rates above 76% but exhibit precision below 45%, even when explicitly instructed to perform minimal debugging.","one_line_summary":"Frontier LLMs pass unit tests over 76% of the time on debugging tasks but achieve edit precision below 45%, indicating regeneration rather than precise debugging.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The automatically synthesized verified atomic bugs and their compositions accurately represent real-world debugging scenarios, and the new edit-level precision and bug-level recall metrics validly measure 'precise debugging' behavior.","pith_extraction_headline":"Frontier LLMs pass debugging tests above 76 percent yet edit with precision below 45 percent, often regenerating entire solutions instead of making targeted fixes."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.17338/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"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"}