{"paper":{"title":"From I/O to Code with Discovery Agent","license":"http://creativecommons.org/licenses/by/4.0/","headline":"DIO-Agent synthesizes code from input-output examples by framing the task as evolutionary search with an LLM mutation operator guided by a simplicity-first bias.","cross_cats":["cs.AI","cs.CL","cs.SE"],"primary_cat":"cs.LG","authors_text":"Binhua Li, Ge Li, Haoran Zhang, Jiaru Qian, Peixu Wang, Xiaokang Yang, Xue Jiang, Yihong Dong, Yongbin Li, Zhi Jin","submitted_at":"2026-05-14T18:57:32Z","abstract_excerpt":"The automatic synthesis of a program from any form of specification is regarded as a holy grail of computer science. Fueled by LLMs, NL2Code has achieved tremendous success, yet the fundamentally more challenging task of synthesizing programs from input-output behavior, which we refer to as IO2Code, remains largely unsolved. Whereas NL2Code can exploit the semantic alignment between natural language and code acquired during pretraining, IO2Code requires recovering underlying principles from concrete computational behavior, navigating a vast and underspecified hypothesis space. To address this,"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"DIO-Agent consistently outperforms both traditional program-by-example method and SOTA evolution-agent baselines across all difficulty levels and various LLMs, while substantially surpassing test-time scaling strategies with equivalent sampling budgets.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The Transformation Priority Premise successfully biases the LLM mutation operator toward the simplest hypothesis consistent with current evidence without missing valid complex solutions or introducing bias that harms search on harder instances.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"DIO-Agent frames IO2Code as LLM-driven evolutionary search over programs with a Transformation Priority Premise to favor simple hypotheses, outperforming baselines on a new IO2CodeBench.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"DIO-Agent synthesizes code from input-output examples by framing the task as evolutionary search with an LLM mutation operator guided by a simplicity-first bias.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"d0a193ea65c90ca5d6a52cbc1b059a759d409a9da9ca846fdb368a5bce88b166"},"source":{"id":"2605.15334","kind":"arxiv","version":1},"verdict":{"id":"d42e2080-0d3d-489b-aa0b-a2c4caf93b6a","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T15:56:33.555461Z","strongest_claim":"DIO-Agent consistently outperforms both traditional program-by-example method and SOTA evolution-agent baselines across all difficulty levels and various LLMs, while substantially surpassing test-time scaling strategies with equivalent sampling budgets.","one_line_summary":"DIO-Agent frames IO2Code as LLM-driven evolutionary search over programs with a Transformation Priority Premise to favor simple hypotheses, outperforming baselines on a new IO2CodeBench.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The Transformation Priority Premise successfully biases the LLM mutation operator toward the simplest hypothesis consistent with current evidence without missing valid complex solutions or introducing bias that harms search on harder instances.","pith_extraction_headline":"DIO-Agent synthesizes code from input-output examples by framing the task as evolutionary search with an LLM mutation operator guided by a simplicity-first bias."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.15334/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-19T16:04:48.843362Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-19T16:01:18.136676Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T14:41:54.189776Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T13:33:22.759757Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"c78afbb3a260a595f5902923a2fc02dd47f8b4091b5a1b59fabfb5f34d4a22be"},"references":{"count":28,"sample":[{"doi":"","year":null,"title":"Codeevolve: An open source evolutionary coding agent for algorithm discovery and optimization","work_id":"b31e9fdb-cb9e-45bb-8e54-e6fc74ca613a","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1901,"title":"Language models are few-shot learners","work_id":"e212f3e7-2f01-4e2b-bffa-27c7cf3727b0","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Magellan: Autonomous discovery of novel compiler optimization heuristics with alphaevolve.arXiv preprint arXiv:2601.21096,","work_id":"68c55fad-65fe-4969-8b39-f7c56fbf692b","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"Evaluating Large Language Models Trained on Code","work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","ref_index":4,"cited_arxiv_id":"2107.03374","is_internal_anchor":true},{"doi":"10.1145/3425898.3426952","year":null,"title":"Yihong Dong, Xue Jiang, Zhi Jin, and Ge Li","work_id":"1cc572f0-5377-495c-abac-e100b48187bf","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":28,"snapshot_sha256":"3d99557407e6fb822a32c00391df803a35fd30e557e2f5734f3ac1e41aaf5cfe","internal_anchors":5},"formal_canon":{"evidence_count":1,"snapshot_sha256":"3efe7107d4e4a787006b7af2d6f0ce0ac43581d0099701fb195372c706586c94"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}