{"work":{"id":"042493e9-b26f-4b4e-bbde-382072ca9b08","openalex_id":"https://openalex.org/W3177813494","doi":"10.48550/arxiv.2107.03374","arxiv_id":"2107.03374","raw_key":null,"title":"Evaluating Large Language Models Trained on Code","authors":null,"authors_text":"Mark Chen et al","year":2021,"venue":"cs.LG","abstract":"We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of our problems with 100 samples per problem. Careful investigation of our model reveals its limitations, including difficulty with docstrings describing long chains of operations and with binding operations to variables. Finally, we discuss the potential broader impacts of deploying powerful code generation technologies, covering safety, security, and economics.","external_url":"https://arxiv.org/abs/2107.03374","cited_by_count":1427,"metadata_source":"pith","metadata_fetched_at":"2026-08-05T02:28:24.338817+00:00","pith_arxiv_id":"2107.03374","created_at":"2026-05-08T17:28:41.986595+00:00","updated_at":"2026-08-05T02:28:24.338817+00:00","title_quality_ok":true,"display_title":"Evaluating Large Language Models Trained on Code","render_title":"Evaluating Large Language Models Trained on Code"},"hub":{"state":{"work_id":"042493e9-b26f-4b4e-bbde-382072ca9b08","tier":"mega_hub","tier_reason":"1,000+ Pith inbound or 100,000+ external citations","pith_inbound_count":1346,"external_cited_by_count":1427,"distinct_field_count":41,"first_pith_cited_at":"2021-05-20T17:58:42+00:00","last_pith_cited_at":"2026-07-09T17:35:02+00:00","author_build_status":"needed","summary_status":"needed","contexts_status":"needed","graph_status":"needed","ask_index_status":"needed","reader_status":"needed","recognition_status":"needed","updated_at":"2026-08-21T02:19:22.473969+00:00","tier_text":"mega_hub"},"tier":"mega_hub","role_counts":[{"context_role":"background","n":161},{"context_role":"dataset","n":64},{"context_role":"method","n":12},{"context_role":"other","n":3},{"context_role":"baseline","n":2}],"polarity_counts":[{"context_polarity":"background","n":157},{"context_polarity":"use_dataset","n":57},{"context_polarity":"use_method","n":12},{"context_polarity":"unclear","n":9},{"context_polarity":"support","n":5},{"context_polarity":"baseline","n":2}],"runs":{"ask_index":{"job_type":"ask_index","status":"succeeded","result":{"title":"Evaluating Large Language Models Trained on Code","claims":[{"claim_text":"We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of ou","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Evaluating Large Language Models Trained on Code because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-13T18:03:33.884013+00:00"},"author_expand":{"job_type":"author_expand","status":"succeeded","result":{"authors_linked":[{"id":"898ecdc3-b47e-4729-b950-fab2383cfa7a","orcid":null,"display_name":"Mark Chen et al"}]},"error":null,"updated_at":"2026-05-13T18:03:33.881753+00:00"},"context_extract":{"job_type":"context_extract","status":"succeeded","result":{"enqueued_papers":25},"error":null,"updated_at":"2026-05-13T18:03:33.779895+00:00"},"graph_features":{"job_type":"graph_features","status":"succeeded","result":{"co_cited":[{"title":"Training Verifiers to Solve Math Word Problems","work_id":"acab1aa8-b4d6-40e0-a3ee-25341701dca2","shared_citers":139},{"title":"Program Synthesis with Large Language Models","work_id":"fd241a05-03b9-4de2-9588-9d77ce176125","shared_citers":134},{"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","shared_citers":86},{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":79},{"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","shared_citers":69},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":61},{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":61},{"title":"Code Llama: Open Foundation Models for Code","work_id":"e73bffa4-7620-47ac-9327-259a60db52ca","shared_citers":57},{"title":"Llama 2: Open Foundation and Fine-Tuned Chat Models","work_id":"68a5177f-d644-44c1-bd4f-4e5278c22f5d","shared_citers":51},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":47},{"title":"Measuring Mathematical Problem Solving With the MATH Dataset","work_id":"50652ac6-fb7c-4675-a2c2-159c241feb17","shared_citers":46},{"title":"Scaling Laws for Neural Language Models","work_id":"b7dd8749-9c45-4977-ab9b-64478dce1ae8","shared_citers":40},{"title":"SWE-bench: Can Language Models Resolve Real-World GitHub Issues?","work_id":"d0effe15-a689-441a-8e3f-ea35f1c4e4b1","shared_citers":40},{"title":"DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence","work_id":"f22dae5a-27e2-41d0-a061-c4286418dee3","shared_citers":36},{"title":"LLaMA: Open and Efficient Foundation Language Models","work_id":"c018fc23-6f3f-4035-9d02-28a2173b2b9d","shared_citers":36},{"title":"Qwen2.5-Coder Technical Report","work_id":"09ba463d-6377-4017-9801-444ffb94b056","shared_citers":35},{"title":"Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge","work_id":"28ea1282-d657-4c61-a83c-f1249be6d6b1","shared_citers":34},{"title":"DeepSeek-V3 Technical Report","work_id":"57d2791d-2219-4c31-a077-afc04b12a75c","shared_citers":33},{"title":"Measuring Massive Multitask Language Understanding","work_id":"e87ec49a-544b-4ec8-8991-75298c64ff5e","shared_citers":33},{"title":"Qwen2.5 Technical Report","work_id":"d8432992-4980-4a81-85c7-9fa2c2b87f85","shared_citers":33},{"title":"LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code","work_id":"ea9e51ce-1e75-4182-92d8-4d25f70d2ee4","shared_citers":31},{"title":"Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters","work_id":"a8d50b24-bdf5-46ed-bc4f-2927dfd81f1d","shared_citers":29},{"title":"DAPO: An Open-Source LLM Reinforcement Learning System at Scale","work_id":"64019d00-0b11-4bbd-b173-b46c8fad0157","shared_citers":28},{"title":"StarCoder: may the source be with you!","work_id":"7e9c3d6e-d6f7-4763-9ef6-de471506c58f","shared_citers":27}],"time_series":[{"n":5,"year":2021},{"n":9,"year":2022},{"n":16,"year":2023},{"n":21,"year":2024},{"n":13,"year":2025},{"n":380,"year":2026}]},"error":null,"updated_at":"2026-05-13T17:25:55.417193+00:00"},"identity_refresh":{"job_type":"identity_refresh","status":"succeeded","result":{"fixed":1,"items":[{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","resolver":"local_arxiv","confidence":0.98,"old_work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e"}],"errors":[],"attempted":1},"error":null,"updated_at":"2026-05-13T18:03:33.147300+00:00"},"reader_index":{"job_type":"reader_index","status":"succeeded","result":{"note":"annotated reader requires full-text/OA fetch; shell is wired for mega hubs","status":"reader queued"},"error":null,"updated_at":"2026-06-30T09:39:42.728017+00:00"},"recognition_alignment":{"job_type":"recognition_alignment","status":"succeeded","result":{"modules":["IndisputableMonolith.StandardModel.SupersymmetryBreaking","IndisputableMonolith.Cognition.AnimalZComplexityBound","IndisputableMonolith.Physics.StandardModelGroupStructure","IndisputableMonolith.Physics.StandardModelLagrangianStructure","IndisputableMonolith.StandardModel.StrongCP","IndisputableMonolith.Cosmology.Inflation","IndisputableMonolith.Patterns.GrayCodeAxioms","IndisputableMonolith.Information.ChurchTuring"],"query_chars":1009},"error":null,"updated_at":"2026-06-30T09:40:04.377210+00:00"},"role_polarity":{"job_type":"role_polarity","status":"succeeded","result":{"title":"Evaluating Large Language Models Trained on Code","claims":[{"claim_text":"We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of ou","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Evaluating Large Language Models Trained on Code because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-13T18:03:33.782984+00:00"},"summary_claims":{"job_type":"summary_claims","status":"succeeded","result":{"title":"Evaluating Large Language Models Trained on Code","claims":[{"claim_text":"We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of ou","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Evaluating Large Language Models Trained on Code because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-13T17:25:52.714606+00:00"}},"summary":{"title":"Evaluating Large Language Models Trained on Code","claims":[{"claim_text":"We introduce Codex, a GPT language model fine-tuned on publicly available code from GitHub, and study its Python code-writing capabilities. A distinct production version of Codex powers GitHub Copilot. On HumanEval, a new evaluation set we release to measure functional correctness for synthesizing programs from docstrings, our model solves 28.8% of the problems, while GPT-3 solves 0% and GPT-J solves 11.4%. Furthermore, we find that repeated sampling from the model is a surprisingly effective strategy for producing working solutions to difficult prompts. Using this method, we solve 70.2% of ou","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks Evaluating Large Language Models Trained on Code because it crossed a citation-hub threshold.","role_counts":[]},"graph":{"co_cited":[{"title":"Training Verifiers to Solve Math Word Problems","work_id":"acab1aa8-b4d6-40e0-a3ee-25341701dca2","shared_citers":139},{"title":"Program Synthesis with Large Language Models","work_id":"fd241a05-03b9-4de2-9588-9d77ce176125","shared_citers":134},{"title":"GPT-4 Technical Report","work_id":"b928e041-6991-4c08-8c81-0359e4097c7b","shared_citers":86},{"title":"Qwen3 Technical Report","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","shared_citers":79},{"title":"The Llama 3 Herd of Models","work_id":"1549a635-88af-4ac1-acfe-51ae7bb53345","shared_citers":69},{"title":"DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models","work_id":"c5006563-f3ec-438a-9e35-b7b484f34828","shared_citers":61},{"title":"DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning","work_id":"e6b75ad5-2877-4168-97c8-710407094d20","shared_citers":61},{"title":"Code Llama: Open Foundation Models for Code","work_id":"e73bffa4-7620-47ac-9327-259a60db52ca","shared_citers":57},{"title":"Llama 2: Open Foundation and Fine-Tuned Chat Models","work_id":"68a5177f-d644-44c1-bd4f-4e5278c22f5d","shared_citers":51},{"title":"Proximal Policy Optimization Algorithms","work_id":"240c67fe-d14d-4520-91c1-38a4e272ca19","shared_citers":47},{"title":"Measuring Mathematical Problem Solving With the MATH Dataset","work_id":"50652ac6-fb7c-4675-a2c2-159c241feb17","shared_citers":46},{"title":"Scaling Laws for Neural Language Models","work_id":"b7dd8749-9c45-4977-ab9b-64478dce1ae8","shared_citers":40},{"title":"SWE-bench: Can Language Models Resolve Real-World GitHub Issues?","work_id":"d0effe15-a689-441a-8e3f-ea35f1c4e4b1","shared_citers":40},{"title":"DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence","work_id":"f22dae5a-27e2-41d0-a061-c4286418dee3","shared_citers":36},{"title":"LLaMA: Open and Efficient Foundation Language Models","work_id":"c018fc23-6f3f-4035-9d02-28a2173b2b9d","shared_citers":36},{"title":"Qwen2.5-Coder Technical Report","work_id":"09ba463d-6377-4017-9801-444ffb94b056","shared_citers":35},{"title":"Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge","work_id":"28ea1282-d657-4c61-a83c-f1249be6d6b1","shared_citers":34},{"title":"DeepSeek-V3 Technical Report","work_id":"57d2791d-2219-4c31-a077-afc04b12a75c","shared_citers":33},{"title":"Measuring Massive Multitask Language Understanding","work_id":"e87ec49a-544b-4ec8-8991-75298c64ff5e","shared_citers":33},{"title":"Qwen2.5 Technical Report","work_id":"d8432992-4980-4a81-85c7-9fa2c2b87f85","shared_citers":33},{"title":"LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code","work_id":"ea9e51ce-1e75-4182-92d8-4d25f70d2ee4","shared_citers":31},{"title":"Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters","work_id":"a8d50b24-bdf5-46ed-bc4f-2927dfd81f1d","shared_citers":29},{"title":"DAPO: An Open-Source LLM Reinforcement Learning System at Scale","work_id":"64019d00-0b11-4bbd-b173-b46c8fad0157","shared_citers":28},{"title":"StarCoder: may the source be with you!","work_id":"7e9c3d6e-d6f7-4763-9ef6-de471506c58f","shared_citers":27}],"time_series":[{"n":5,"year":2021},{"n":9,"year":2022},{"n":16,"year":2023},{"n":21,"year":2024},{"n":13,"year":2025},{"n":380,"year":2026}]},"authors":[{"id":"898ecdc3-b47e-4729-b950-fab2383cfa7a","orcid":null,"display_name":"Mark Chen et al","source":"manual","import_confidence":0.72}]},"citers":{"total":1346,"items":[{"citing_arxiv_id":"2607.08734","ref_index":91,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"The Illusion of Equivalency: Statistical Characterization of Quantization Effects in LLMs","primary_cat":"cs.AI","submitted_at":"2026-07-09T17:35:02+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Quantized LLMs diverge from their base models at the decision level even when accuracy is preserved, with query and key attention projections showing the greatest structural distortion under low-bit compression.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.08691","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"ProjAgent: Procedural Similarity Retrieval for Repository-Level Code Generation","primary_cat":"cs.SE","submitted_at":"2026-07-09T16:50:54+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"ProjAgent introduces procedural similarity—retrieving code with matching computational logic—via LLM hidden-state projections, improving repository-level code generation to 41.14% Pass@1 on REPOCOD.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.08221","ref_index":1,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"LUMI: Tokenizer-Agnostic LLM-Based Lossless Image Compression","primary_cat":"cs.CV","submitted_at":"2026-07-09T08:16:14+00:00","verdict":"CONDITIONAL","verdict_confidence":"HIGH","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A tokenizer-free pixel embedding, position encoding, and 256-way head let frozen LLMs act as portable entropy models for lossless RGB compression across model families.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.08124","ref_index":47,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"TTHE: Test-Time Harness Evolution","primary_cat":"cs.SE","submitted_at":"2026-07-09T05:53:39+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"An LLM agent can improve itself at test time by rewriting its surrounding executable harness from unlabeled traces, using only proxy signals and a frozen model.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.08009","ref_index":29,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"From Execution to Education: A Bloom-Aligned Framework for Measuring Educational Control in LLMs","primary_cat":"cs.CL","submitted_at":"2026-07-09T00:27:07+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.5,"formal_verification":"none","one_line_summary":"On 2,520 programming tasks, matched Qwen general and coder models reliably raise Bloom cognitive demand but fail to lower it, so execution skill does not imply educational control.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.07946","ref_index":2,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"DeepSWE: Measuring Frontier Coding Agents on Original, Long-Horizon Engineering Tasks","primary_cat":"cs.SE","submitted_at":"2026-07-08T21:45:34+00:00","verdict":"CONDITIONAL","verdict_confidence":"HIGH","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Original, never-upstreamed multi-file engineering tasks with functional verifiers grade coding agents more faithfully and separate frontier models more widely than inherited-test SWE benchmarks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.07881","ref_index":1,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Functional and Secure Code Generation with Task Vectors","primary_cat":"cs.SE","submitted_at":"2026-07-08T19:32:22+00:00","verdict":"ACCEPT","verdict_confidence":"HIGH","novelty_score":6.0,"formal_verification":"none","one_line_summary":"LPO-derived Secure-Anchored task vectors raise simultaneous functional-and-secure code rates by 2.1–36 pp on six coding LLMs with near-zero inference overhead.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.07702","ref_index":1,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"From Noisy Traces to Root Causes: Structural Trajectory Analysis and Causal Extraction for Agent Optimization","primary_cat":"cs.CL","submitted_at":"2026-07-08T17:57:27+00:00","verdict":"CONDITIONAL","verdict_confidence":"UNKNOWN","novelty_score":6.0,"formal_verification":"none","one_line_summary":"STRACE improves long-horizon agent optimization by constructing a dependency graph to filter redundant traces and localize true root-cause modules for targeted policy updates.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.07779","ref_index":35,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"From Solvers to Research: Large Language Model-Driven Formal Mathematics at the Research Frontier","primary_cat":"cs.CL","submitted_at":"2026-07-08T17:46:36+00:00","verdict":"ACCEPT","verdict_confidence":"HIGH","novelty_score":6.0,"formal_verification":"none","one_line_summary":"LLM formal provers must shift from competition solvers to research agents that handle open-ended, under-specified frontier mathematics under machine-checked rigor.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.07619","ref_index":103,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Rethinking Code Performance Benchmarks for LLMs","primary_cat":"cs.SE","submitted_at":"2026-07-08T16:36:03+00:00","verdict":"CONDITIONAL","verdict_confidence":"HIGH","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Re-evaluating four LLM code-efficiency benchmarks with 30-run statistical testing shows 93.89% of 'performant' implementations are indistinguishable from baselines; a multi-agent test-generation framework reveals hidden significant improvements in ~24% of previously non-significant tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.07554","ref_index":27,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"RubriQ: Rubric-Guided Group Relative Policy Optimization for Constraint-Aware Quantum Circuit Synthesis","primary_cat":"quant-ph","submitted_at":"2026-07-08T15:50:54+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A rubric-guided GRPO pipeline fine-tunes a 7B LLM to synthesize quantum circuits achieving 3.31x T-gate compression with <1% hardware-constraint violations, validated on IBM and IonQ processors.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.07480","ref_index":10,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Biased or Personalized? The Impact of Personal Information on AI-driven Development","primary_cat":"cs.SE","submitted_at":"2026-07-08T14:43:08+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Changing only the prompter's age and gender in AI coding prompts produces statistically significant differences in generated website interface design, template content, and code structure across 800 generated websites and a 20-person user study.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.07409","ref_index":9,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"DeLS-Spec: Decoupled Long-Short Contexts for Parallel Speculative Drafting","primary_cat":"cs.CL","submitted_at":"2026-07-08T13:41:52+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":5.0,"formal_verification":"none","one_line_summary":"DeLS-Spec improves block-parallel speculative decoding by fusing DFlash logits with an independently trained lightweight local head and a unigram prior correction.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.07321","ref_index":4,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"From Atomic Actions to Standard Operating Procedures: Iterative Tool Optimization for Self-Evolving LLM Agents","primary_cat":"cs.AI","submitted_at":"2026-07-08T12:09:49+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":5.0,"formal_verification":"none","one_line_summary":"EvoSOP enables LLM agents to self-evolve by iteratively synthesizing atomic tool actions into reusable Standard Operating Procedures, improving success rates and reducing reasoning rounds on ACEBench and Tau2Bench.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.07748","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Selective Left-Shift: Turning Test-Time Compute and Difficulty-based Curation into Training Data for Low-Resource Code Generation","primary_cat":"cs.LG","submitted_at":"2026-07-08T09:18:56+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Left-shifting iterative compiler/test refinement into verified SFT data, then GRPO on difficulty-curated IO rewards, lifts Qwen3-8B Julia pass@1 past prior SOTA at 1/3 data and 1/6 cost, and bootstraps Ballerina.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.07744","ref_index":1,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"PERFOPT-Bench: Evaluating Coding Agents on Software Performance Optimization","primary_cat":"cs.SE","submitted_at":"2026-07-08T08:32:18+00:00","verdict":"CONDITIONAL","verdict_confidence":"HIGH","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Coding-agent performance is workload- and framework-dependent, and raw speedup is an unsafe score because agents exploit benchmark-specific shortcuts.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.06764","ref_index":10,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Cost-Effective Agent Harnesses for Abstract Reasoning and Generalization on ARC-AGI-1","primary_cat":"cs.AI","submitted_at":"2026-07-07T19:49:35+00:00","verdict":"CONDITIONAL","verdict_confidence":"HIGH","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Explorer-Definer and Reflective Orchestrator harnesses raise DeepSeek V3.2 from 15.5% to 67.25% pass@2 on ARC-AGI-1 public eval at $0.25–$0.62 per task without ARC-specific training.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.06721","ref_index":11,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Flowcode: An AI-Powered Programming Environment for Scaffolding Iteration in Creative Computing Education","primary_cat":"cs.HC","submitted_at":"2026-07-07T18:39:39+00:00","verdict":"CONDITIONAL","verdict_confidence":"HIGH","novelty_score":6.0,"formal_verification":"none","one_line_summary":"An iterative design of Flowcode shows that LLM-generated flowcharts plus step-by-step fill-in-the-blank explanations help novices orient, navigate, and extend found creative-coding projects.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.06720","ref_index":20,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"When Does In-Context Search Help? A Sampling-Complexity Theory of Reflection-Driven Reasoning","primary_cat":"cs.AI","submitted_at":"2026-07-07T18:36:04+00:00","verdict":"CONDITIONAL","verdict_confidence":"HIGH","novelty_score":7.0,"formal_verification":"none","one_line_summary":"When reflections localize early errors, in-context search solves exp-small pass-rate problems with poly sequential attempts; otherwise it offers no asymptotic gain over parallel sampling, and the update is learnable and RLVR-optimal.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.06411","ref_index":15,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"RuBench: A Repository-Level Agentic Coding Benchmark with Natively Authored Russian Task Specifications","primary_cat":"cs.SE","submitted_at":"2026-07-07T15:41:22+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":8.0,"formal_verification":"none","one_line_summary":"RuBench, the first repository-level coding benchmark with natively authored (non-translated) Russian task specifications, measures deployed coding agents on 25 contamination-gated fix tasks and documents model substitution and answer leakage that change how agent benchmarks must be audited.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.06636","ref_index":4,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Specification Grounding Drives Test Effectiveness for LLM Code","primary_cat":"cs.SE","submitted_at":"2026-07-07T14:11:00+00:00","verdict":"CONDITIONAL","verdict_confidence":"HIGH","novelty_score":6.5,"formal_verification":"none","one_line_summary":"Specification grounding, not test quantity, is the main reason AI-written tests improve LLM code: +38 pp correctness and 0% false alarms over a strong edge-prompted baseline.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.06125","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Evaluating Fine-Tuning and Metrics for Neural Decompilation of Dart AOT Binaries","primary_cat":"cs.SE","submitted_at":"2026-07-07T10:36:12+00:00","verdict":"ACCEPT","verdict_confidence":"HIGH","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Fine-tuning small LLMs for Dart decompilation yields no functional improvement and surface metrics can diverge from correctness.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.06088","ref_index":2,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Flow Matching-Based Speech Source Separation with Best-of-N Biometric Sampling","primary_cat":"cs.SD","submitted_at":"2026-07-07T10:00:23+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":5.0,"formal_verification":"none","one_line_summary":"A flow-matching speech separator with biometric best-of-N candidate selection and chunk-wise channel alignment achieves competitive separation metrics and the best downstream ASR/SV error rates among evaluated systems on Libri2Mix.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.06009","ref_index":4,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Multi-Channel Spread-Spectrum Code Watermarking","primary_cat":"cs.CR","submitted_at":"2026-07-07T08:50:19+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"A training-free post-hoc code watermark embeds 24-bit identifiers via multi-channel spread-spectrum encoding over naming conventions and semantic pattern pairs, with majority voting and Reed-Solomon recovery.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.06004","ref_index":15,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Large Language Models Have Unreliable Understanding of Software Engineering Terminology","primary_cat":"cs.SE","submitted_at":"2026-07-07T08:46:49+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"LLMs detect falsified ISO/IEC/IEEE 24765 SE definitions well but also reject many correct ones, indicating systematic rejection bias rather than genuine discriminative understanding of SE terminology.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.05992","ref_index":14,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"PluraMath: Extending Mathematical Reasoning Evaluation Beyond High-Resource Languages","primary_cat":"cs.CL","submitted_at":"2026-07-07T08:25:29+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"PluraMath extends PolyMath with human-validated math problems in 18 mid-to-extreme low-resource languages and benchmarks 27 reasoning LLMs, finding a persistent high- vs low-resource performance gap.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.05842","ref_index":1,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Beyond Refusal: A Same-Lineage Study of Aligned and Abliterated LLMs for Vulnerability Analysis","primary_cat":"cs.SE","submitted_at":"2026-07-07T05:04:48+00:00","verdict":"CONDITIONAL","verdict_confidence":"UNKNOWN","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Refusal-ablated LLMs outperform aligned models on code-grounded localization and early executable patch generation, while aligned models retain advantages on shallow diagnostic tasks under neutral wording.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.05762","ref_index":7,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Articulating Assumptions in AI-Generated Scientific Analyses through Task Decomposition","primary_cat":"cs.SE","submitted_at":"2026-07-07T02:40:32+00:00","verdict":"CONDITIONAL","verdict_confidence":"HIGH","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Quantity-grounded multi-agent decomposition makes LLM-generated collider analysis code inspectable and reliable with 14B-scale models, outperforming prior single-prompt approaches.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.05717","ref_index":5,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Plainbook: Data Science, in Plain Language","primary_cat":"cs.HC","submitted_at":"2026-07-07T01:00:54+00:00","verdict":"CONDITIONAL","verdict_confidence":"HIGH","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Plainbook makes data-science notebooks natural-language-first by preserving cell descriptions, generating code via AI, enforcing linear execution via a checkpointing kernel, and adding value-centered cell and global tests.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.05708","ref_index":7,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Akashic: A Low-Overhead LLM Inference Service with MemAttention","primary_cat":"cs.AI","submitted_at":"2026-07-07T00:06:22+00:00","verdict":"CONDITIONAL","verdict_confidence":"HIGH","novelty_score":5.5,"formal_verification":"none","one_line_summary":"Akashic’s MemAttention plus locality-aware placement improves agent task accuracy by up to 10.2 points and throughput by up to 1.21× over prior memory systems across four long-horizon workloads.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.05704","ref_index":4,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"LLM-Driven Neural Network Generation with Same-Family Architecture Guidance: Disentangling Transfer and Adaptation","primary_cat":"cs.LG","submitted_at":"2026-07-06T23:59:18+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Same-family source guidance lets LLMs improve weak neural nets over non-source controls, with recipe adaptation (not mere copy) explaining large SVHN AlexNet gains.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.05197","ref_index":3,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Is Three the Magic Number? An Empirical Evaluation of LLM-Based Repair Loops","primary_cat":"cs.SE","submitted_at":"2026-07-06T15:14:10+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Across five LLM-based SE tools and three models, 3-4 repair iterations capture most achievable gains, with diminishing returns thereafter.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.05196","ref_index":129,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Unified Audio Intelligence Without Regressing on Text Intelligence","primary_cat":"cs.CL","submitted_at":"2026-07-06T15:11:57+00:00","verdict":"CONDITIONAL","verdict_confidence":"HIGH","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Audex unifies audio understanding and generation on a strong text MoE backbone with multi-stage SFT plus text-only Cascade RL, matching open SOTA audio scores while mostly retaining text capability.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.02464","ref_index":3,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Will Scaling Improve Social Simulation with LLMs?","primary_cat":"cs.CL","submitted_at":"2026-07-02T17:30:38+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Using 85 controlled and 35 public LLMs, the authors show social-simulation accuracy generally improves with compute, but some behavioral and low-resource tasks do not scale.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.02390","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"DecompRL: Solving Harder Problems by Learning Modular Code Generation","primary_cat":"cs.LG","submitted_at":"2026-07-02T16:25:10+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"DecompRL is an RL method that learns modular code decomposition for LLMs, enabling exponential candidate generation via recombination to solve harder coding problems with lower GPU cost.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.02333","ref_index":55,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Guiding Human Validation of LLM-Generated Code via Verifiable Literate Programming","primary_cat":"cs.SE","submitted_at":"2026-07-02T15:37:54+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"VLP adds an NL documentation layer with trace-linked mismatch detection and derived formal checks to make human validation of LLM code feasible, lifting pass@1 from 28.7-73.2% to 65.4-93.5%.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.02329","ref_index":1,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Grounded autonomous research: a fault-tolerant LLM pipeline from corpus to manuscript in frontier computational physics","primary_cat":"cs.AI","submitted_at":"2026-07-02T15:35:41+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"An LLM pipeline with fresh-context sessions and literature calibration produces a publication-grade manuscript with three substantive findings on altermagnetic piezomagnetism from a corpus of 11,083 papers.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.02186","ref_index":29,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"UA-ChatDev: Uncertainty-Aware Multi-Agent Collaboration for Reliable Software Development","primary_cat":"cs.AI","submitted_at":"2026-07-02T13:56:57+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"UA-ChatDev integrates token-level uncertainty estimation and phase-aware verification into multi-agent software development and reports better benchmark scores than prior frameworks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.02118","ref_index":54,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Enhancing Fitness Intelligence through Domain-Specific LLM Post-Training","primary_cat":"cs.AI","submitted_at":"2026-07-02T12:53:41+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":3.0,"formal_verification":"none","one_line_summary":"FitOne-8B/32B models improve average scores on ACSM-EP and NSCA-CSCS certification exams by up to 12.73% over base Qwen3 while retaining general capabilities.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.02057","ref_index":8,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Prompt Coverage Adequacy","primary_cat":"cs.SE","submitted_at":"2026-07-02T11:35:39+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Prompt Coverage Adequacy, measured via attention boosting in LLMs, is associated with fault detection and uncovers over 30% more faults than traditional code coverage when guiding test generation across two datasets.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.02032","ref_index":41,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"PACE: A Proxy for Agentic Capability Evaluation","primary_cat":"cs.AI","submitted_at":"2026-07-02T10:59:03+00:00","verdict":"CONDITIONAL","verdict_confidence":"HIGH","novelty_score":6.0,"formal_verification":"none","one_line_summary":"A dual local/global selection of ~100 non-agentic instances predicts four agentic benchmarks with LOOCV MAE under 4%, Spearman above 0.80, and ~85% pairwise ranking accuracy at under 1% of full agent cost.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01953","ref_index":7,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Underspecification does not imply Incoherence: The Risks of Semantic Collapse in Coding Models","primary_cat":"cs.SE","submitted_at":"2026-07-02T09:43:15+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Coding LLMs exhibit detrimental semantic collapse on underspecified prompts by producing consistent but incorrect code rather than incoherent variations, affecting 3-32% of tasks across MBPP, HumanEval, and LiveCodeBench.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01927","ref_index":20,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"TUDUM: A Turkish-Thinking Reasoning Pipeline for Qwen3.5-27B","primary_cat":"cs.CL","submitted_at":"2026-07-02T09:22:36+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":3.0,"formal_verification":"none","one_line_summary":"TUDUM applies LoRA-based SFT on 15,991 Turkish reasoning examples followed by GRPO reinforcement learning on Turkish math problems to a 27B Qwen model, producing shorter Turkish reasoning traces with mixed benchmark results.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01883","ref_index":77,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"PairCoder++: Pair Programming as a Universal Paradigm for Verified Code-Driven Multimodal and Structured-Artifact Generation","primary_cat":"cs.CL","submitted_at":"2026-07-02T08:36:02+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"PairCoder is a two-agent pair-programming method that leverages toolchain verification oracles to improve LLM generation of verifiable structured artifacts on 17 benchmarks across seven models.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01867","ref_index":62,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"An Exploratory Study on LLM-Generated Code and Comments in Code Repositories","primary_cat":"cs.SE","submitted_at":"2026-07-02T08:25:15+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Detector proxy analysis of real repositories finds LLM-like code decreasing over time and common in tests, with more in company repos, substantial clones, stable but often ungrammatical comments, and few linked bugs.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01855","ref_index":9,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Regression Accumulation in Multi-Turn LLM Programming Conversations","primary_cat":"cs.SE","submitted_at":"2026-07-02T08:15:40+00:00","verdict":"CONDITIONAL","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Regression accumulation affects 40-73% of 8-turn LLM coding tasks on extended HumanEval+/MBPP+ benchmarks, with verification gates improving final-turn pass rates on prior tests.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01810","ref_index":4,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Decoupling Code Complexity from Newcomer Participation: A Causal Study of AI Coding Agent Adoption in OSS","primary_cat":"cs.SE","submitted_at":"2026-07-02T07:24:20+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"AI coding agent adoption in OSS projects raises code complexity modestly but produces no causal reduction in newcomer participation per DiD estimates on matched GitHub projects.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01789","ref_index":5,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"EPnG: Adaptive Expert Prune-and-Grow for Parameter-Efficient MoE Fine-tuning","primary_cat":"cs.LG","submitted_at":"2026-07-02T07:02:44+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"EPnG reallocates LoRA capacity in MoE models by pruning experts with low router gate probabilities and expanding high-importance ones via rank growth, outperforming standard LoRA and nearing full fine-tuning performance with 0.55-0.72% parameters updated.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01678","ref_index":2,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"SCAPE: Accurate and Efficient LLM Training with Extreme Sparse Communication","primary_cat":"cs.LG","submitted_at":"2026-07-02T04:10:42+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"SCAPE enables 90-99% sparse gradient communication in sharded Adam-style LLM training by deriving masks from first-moment statistics, achieving up to 43.3% faster pre-training on Llama-500M with no loss in validation loss or downstream accuracy.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01531","ref_index":52,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"OPINE-World: Programmatic World Modeling with Ontology-error-Prioritized Interactive Exploration for ARC-AGI-3","primary_cat":"cs.AI","submitted_at":"2026-07-01T23:04:47+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"An object-centric, LLM-synthesized world model verified by exact replay solves 20 of 25 ARC-AGI-3 games with no per-game training, scoring 78.4 in action efficiency.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01490","ref_index":10,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Don't Let Gains FADE: Breaking Down Policy Gradient Weights in RL","primary_cat":"cs.LG","submitted_at":"2026-07-01T21:39:19+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"FADE is a self-adapting advantage for policy-gradient RL that reads training dynamics to balance positive/negative gradient mass and difficulty focus, yielding faster peak performance and better accuracy-diversity trade-offs than static baselines on LLM reasoning benchmarks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01444","ref_index":1,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"On the Utility and Factual Reliability of Pruned Mixture-of-Experts Models in the Biomedical Domain","primary_cat":"cs.LG","submitted_at":"2026-07-01T20:08:19+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"Moderate pruning of MoE models preserves in-domain biomedical utility and reliability but both degrade rapidly in cross-domain settings and at extreme pruning ratios.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01360","ref_index":1,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Benchmarking Code Improvement with Progressive, Adaptive, and Interactive Feedback","primary_cat":"cs.SE","submitted_at":"2026-07-01T18:20:27+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"PAIR-Bench defines a progressive hinting protocol with failure-region and hint-depth controls to measure LLM code refinement trajectories in detail.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01213","ref_index":10,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"RepoRescue: An Empirical Study of LLM Agents on Whole-Repository Compatibility Rescue","primary_cat":"cs.SE","submitted_at":"2026-07-01T17:51:28+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"RepoRescue creates a benchmark of 315 repositories and shows LLM agents rescue up to 41.5% with runtime enforcement and 62.7% when combining systems, with hardest cases requiring cross-file changes.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01211","ref_index":20,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Are Performance-Optimization Benchmarks Reliably Measuring Coding Agents?","primary_cat":"cs.SE","submitted_at":"2026-07-01T17:50:48+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Auditing three performance-optimization benchmarks shows reference-patch validity collapses under cross-machine replay, rankings depend on scoring rules, and 384 of 450 replay-valid tasks are already solved by at least one public submission.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.01179","ref_index":2,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"QuasiMoTTo: Quasi-Monte Carlo Test-Time Scaling","primary_cat":"cs.LG","submitted_at":"2026-07-01T17:10:08+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"QuasiMoTTo uses quasi-Monte Carlo to produce correlated yet marginally correct samples from language models, matching i.i.d. pass@k with 25-47% fewer samples on reasoning benchmarks and 50% fewer RL training steps.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00939","ref_index":8,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Leveraging LLM-Based Agentic Systems to Generate Quantum Applications for Test Optimization","primary_cat":"cs.SE","submitted_at":"2026-07-01T13:40:51+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":3.0,"formal_verification":"none","one_line_summary":"QPipe deploys specialized LLM agents for parsing, formulation, code generation, review, execution and verification to produce quantum applications from 20 natural-language test-optimization requirements, reporting 100% compilation and 96.7% execution success with solutions that beat a genetic-algori","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00911","ref_index":16,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"From Registry to Repository: How AI Agent Skills Are Written, Adapted, and Maintained","primary_cat":"cs.SE","submitted_at":"2026-07-01T13:14:45+00:00","verdict":"ACCEPT","verdict_confidence":"HIGH","novelty_score":6.5,"formal_verification":"none","one_line_summary":"AI agent skills are reused mainly as one-time near-verbatim copies; half stay unmodified, later maintenance is additive and local, and the behavioural contract stays almost untouched.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00908","ref_index":1,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Beyond Activation Alignment:The Alignment-Diversity Tradeoff in Task-Aware LLM Quantization","primary_cat":"cs.LG","submitted_at":"2026-07-01T13:12:21+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":7.0,"formal_verification":"none","one_line_summary":"TASA improves task-aware mixed-precision LLM quantization by searching calibration data mixtures via gradient-trace alignment and aggregating perplexity plus reasoning sensitivity signals, enabling 3.5-bit models to match or beat 4-bit baselines with over 20-point gains on GSM8K.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00711","ref_index":5,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"ClarifyCodeBench: Evaluating LLMs on Clarifying Ambiguous Requirements for Code Generation","primary_cat":"cs.SE","submitted_at":"2026-07-01T09:58:59+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"ClarifyCodeBench is a new benchmark with manual annotations and two metrics showing that LLMs strong at code generation are weak at clarifying ambiguous requirements, with performance worsening as ambiguity density rises.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00664","ref_index":32,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"YOMI-Bench: A Benchmark for Evaluating Kanji Reading and Phonological Understanding of LLMs for Japanese","primary_cat":"cs.CL","submitted_at":"2026-07-01T09:13:19+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"YOMI-Bench is a new benchmark of four tasks for kanji reading and phonological understanding in LLMs, showing low performance even for Japanese-specific and commercial models.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00604","ref_index":3,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Vehicle Routing Problem Meets Large Language Models: An Overview and Perspectives","primary_cat":"math.OC","submitted_at":"2026-07-01T08:30:20+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"Survey organizing LLM uses for VRP into modeler, designer, and coordinator roles, covering variants, solvers, benchmarks, and two experiments.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00572","ref_index":52,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"HARC: Coupling Harmfulness and Refusal Directions for Robust Safety Alignment","primary_cat":"cs.AI","submitted_at":"2026-07-01T07:58:16+00:00","verdict":"ACCEPT","verdict_confidence":"HIGH","novelty_score":7.0,"formal_verification":"none","one_line_summary":"HARC couples harmfulness and refusal directions at prompt and response positions, yielding the best robustness-capability-usability trade-off among major safety methods.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00531","ref_index":9,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Active-GRPO: Adaptive Imitation and Self-Improving Reasoning for Molecular Optimization","primary_cat":"cs.LG","submitted_at":"2026-07-01T07:22:46+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Active-GRPO reaches 0.1773 average SRxSim on TOMG-Bench MOLOPT by adaptively switching between imitation and self-reinforcement while upgrading references, outperforming GRPO and RePO.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00466","ref_index":4,"ref_count":2,"confidence":0.98,"is_internal_anchor":true,"paper_title":"ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving","primary_cat":"cs.DC","submitted_at":"2026-07-01T05:34:38+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"ELDR reduces median TPOT by 5.9-13.9% in PD-disaggregated MoE serving via expert signatures from prefill, K-means partitioning, and locality-band routing with KV-co-indexed signature cache.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00341","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"DiscoLoop: Looping Discrete Embeddings and Continuous Hidden States for Multi-hop Reasoning","primary_cat":"cs.CL","submitted_at":"2026-07-01T02:32:02+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":7.0,"formal_verification":"none","one_line_summary":"DiscoLoop adds a decoded token-embedding channel to looped transformers, fixing a representation mismatch that limited implicit multi-hop reasoning and improving OOD generalization.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00162","ref_index":66,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"FRAME: Learning the Adaptation Domain with a Mixture of Fractional-Fourier Experts","primary_cat":"cs.LG","submitted_at":"2026-06-30T20:39:55+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"FRAME adds a learnable fractional-Fourier order per expert in a MoE-LoRA setup so that low-rank updates are placed in the domain where they are most compact, yielding gains over fixed-domain baselines on LLaMA-3.1-8B and Qwen2.5-7B.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00107","ref_index":33,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"The Illusion of Safety: Multi-Tier Verification of AI vs. Human C++ Code","primary_cat":"cs.SE","submitted_at":"2026-06-30T19:48:35+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"AI-generated C++ programs trigger confirmed runtime (sanitizer) violations at roughly twice the odds of human contest solutions, a gap static analysis does not reveal.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.32007","ref_index":24,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"AxDafny: Agentic Verified Code Generation in Dafny","primary_cat":"cs.AI","submitted_at":"2026-06-30T17:39:43+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"AxDafny achieves 92.7% verification success on DafnyBench (6.5 points above prior proof-hint baselines) via verifier-guided repair and introduces the LCB-Pro-Dafny benchmark of 250 problems.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.31717","ref_index":9,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Nonlinearity-Aware LoRA: Structured Gate Adaptation under Low-Rank Constraints","primary_cat":"cs.LG","submitted_at":"2026-06-30T14:21:51+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"NA-LoRA introduces derivative-based temporal-importance masks and activation-specific step scaling to LoRA to reduce selection misalignment in self-gated FFNs, with reported gains on language and vision-language fine-tuning tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.31706","ref_index":3,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"AdaTrans: Automated C to Rust Transformation via Error-Adaptive Repair","primary_cat":"cs.SE","submitted_at":"2026-06-30T14:11:54+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"AdaTrans uses strategy-driven RAG, error-stratified transformation, and multi-stage validation to reach 95.51% mean compilation pass rate and 81.09% solve rate on 104 algorithmic problems with only 1.19% unsafe files.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.31551","ref_index":13,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"AutoTrainess: Teaching Language Models to Improve Language Models Autonomously","primary_cat":"cs.CL","submitted_at":"2026-06-30T12:09:51+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"AutoTrainess exposes training operations via agent-computer interfaces and outperforms CLI-only baselines on PostTrainBench with scores of 26.94 vs 23.21 for GPT-5.4 and similar gains on other models.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.31511","ref_index":5,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Falsification, Not Exposure: An Internally Preregistered Placebo-Controlled Decomposition of Self-Repair Feedback in Frozen Small Code Models","primary_cat":"cs.SE","submitted_at":"2026-06-30T11:26:14+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Preregistered placebo-controlled decomposition shows external executable counterexamples drive self-repair gains in small code models more than re-exposure or self-critique.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00062","ref_index":5,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"AlgoBench: Benchmarking Algorithmic Adaptation in Code Generation","primary_cat":"cs.SE","submitted_at":"2026-06-30T10:13:12+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"AlgoBench creates traceable variants of competitive programming problems via constraint shifts that invalidate original algorithms, paired with complexity metrics that reveal LLMs often produce functionally correct but asymptotically unsuitable solutions.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.31315","ref_index":10,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"BlockPilot: Instance-Adaptive Policy Learning for Diffusion-based Speculative Decoding","primary_cat":"cs.CL","submitted_at":"2026-06-30T08:24:05+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"BlockPilot is an instance-adaptive policy that predicts optimal block size from the prefilling representation for diffusion speculative decoding, reporting 5.92 acceptance length and 4.20x speedup on Qwen3-4B.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.31159","ref_index":9,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"An Empirical Study of Security Calibration in Large Language Models for Code","primary_cat":"cs.SE","submitted_at":"2026-06-30T05:37:39+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Empirical evaluation of three LLMs finds prevalent overconfidence in insecure code generation, with security calibration outperforming functional calibration but both degrading in repository-level settings.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.31121","ref_index":4,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"The Past Is Prologue: A Plug-in Controller for Selective Updates in Sequentially Evolving LLM Memory","primary_cat":"cs.AI","submitted_at":"2026-06-30T04:33:18+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Janus is a method-agnostic plug-in that uses a Memory Momentum Trigger and compact hybrid evaluation to selectively accept LLM memory updates, yielding +2.7 to +4.6 accuracy gains over base updaters on six datasets.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2607.00053","ref_index":10,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"SWE-Router: Routing in Multi-turn Agentic Software Engineering Tasks","primary_cat":"cs.SE","submitted_at":"2026-06-30T01:46:26+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"SWE-Router introduces trajectory-conditioned value-based routing for LLM agents on SWE tasks, with a Bayes-optimality theorem and empirical cost savings while retaining most strong-model performance.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.30810","ref_index":22,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Towards Knowledge Alignment in Code LLMs: Contrastive Unlearning for Evolving APIs","primary_cat":"cs.SE","submitted_at":"2026-06-29T18:34:55+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"CURE applies contrastive unlearning to reduce deprecated API usage in code LLMs and improve correct replacements on a benchmark dataset while preserving general performance.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.30602","ref_index":19,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"MESA: Prioritizing Vulnerable Communication Channels for Securing Multi-Agent Systems","primary_cat":"cs.CR","submitted_at":"2026-06-29T17:40:45+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"MESA ranks MAS communication edges by vulnerability via graph-theoretic metrics and dynamic probes, achieving mean Spearman ρ=+0.60 correlation with empirical per-edge attack success and 3x interception gain when monitoring the top 10%.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.30296","ref_index":1,"ref_count":2,"confidence":0.98,"is_internal_anchor":true,"paper_title":"ManimAgent: Self-Evolving Multimodal Agents for Visual Education","primary_cat":"cs.AI","submitted_at":"2026-06-29T13:37:56+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"ManimAgent improves Manim animation code generation by maintaining a self-growing dual-channel episodic memory of validated successes and failures derived entirely from its own task stream.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.30109","ref_index":10,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"TacEvo: Self-Evolving Architecture Discovery for Robotic Tactile Perception via LLM-Driven Quality-Diversity Search","primary_cat":"cs.RO","submitted_at":"2026-06-29T10:45:01+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"TacEvo is an LLM-driven self-evolving search method that discovers neural architectures for robotic tactile force regression and grating classification, reporting fitness gains of 56.1% and 96.1% over 20 generations.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29999","ref_index":9,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"AlgoSkill: Learning to Design Algorithms by Scheduling Human-Like Skills","primary_cat":"cs.AI","submitted_at":"2026-06-29T09:09:16+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"AlgoSkill improves LLM algorithm design on programming benchmarks by framing it as verification-guided scheduling over a typed skill library with MCTS, outperforming direct generation and self-refinement.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29982","ref_index":8,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Beyond Uniform Experts: Cost-Aware Expert Execution for Efficient Multi-Device MoE Inference","primary_cat":"cs.DC","submitted_at":"2026-06-29T08:57:59+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"CAEE reduces MoE inference latency 8-18% on 671B DeepSeek-R1 by cost-aware expert pruning and low-overhead compensation while keeping accuracy drop under 1%.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29957","ref_index":1,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"SWE-Together: Evaluating Coding Agents in Interactive User Sessions","primary_cat":"cs.SE","submitted_at":"2026-06-29T08:35:15+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Introduces SWE-Together benchmark from 109 real repository tasks, using an LLM user simulator to evaluate coding agents on success rate and corrective turns needed.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29823","ref_index":7,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Experience Graphs: The Data Foundation for Self-Improving Agents","primary_cat":"cs.DB","submitted_at":"2026-06-29T06:02:20+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Trellis treats agent experience graphs as first-class database state so that search patterns become queries, enabling crash recovery, scaling, and closed-loop training as architectural byproducts.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29815","ref_index":42,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"SrDetection: A Self-Referential Framework for Data Leakage Detection in Code Large Language Models","primary_cat":"cs.CL","submitted_at":"2026-06-29T05:48:42+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"SrDetection detects data leakage in Code LLMs via contrast between original benchmark samples and their semantic variants, reporting F1 gains of 21.52 (gray-box) and 14.46 (black-box) over baselines in a controlled testbed.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29702","ref_index":22,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Verified residual-specific explicit derivative kernels for physics-informed learning and discretized PDE adjoints","primary_cat":"physics.comp-ph","submitted_at":"2026-06-29T02:09:56+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Develops and numerically verifies residual-specific explicit derivative kernels that achieve floating-point agreement with AD while delivering 2-4x speedups and lower memory use in PINN training and CFD adjoint workflows.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.30689","ref_index":9,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Citation Discipline in Spec-Driven Development: A Cross-Model Empirical Study of Output Determinism and Automated Hallucination Detection in LLM-Generated Code","primary_cat":"cs.SE","submitted_at":"2026-06-28T19:38:55+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Mandatory per-line citations in SDD frameworks reduce LLM output determinism but enable reliable automated hallucination detection (TDR 86-88%, FPR 0%), a trade-off replicated across Claude and GLM models.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29538","ref_index":1,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"RESOURCE2SKILL: Distilling Executable Agent Skills from Human-Created Multimodal Resources","primary_cat":"cs.SE","submitted_at":"2026-06-28T17:59:21+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":7.0,"formal_verification":"none","one_line_summary":"A multimodal resource-to-skill pipeline converts tutorial videos and other human-created materials into a hierarchical skill wiki, improving software-agent artifact quality by an average of 11.9 points across seven authoring domains.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29301","ref_index":4,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Pointer-CAD v2: Plan-Then-Construct CAD Generation with Dimension-Aware Parametric Precision","primary_cat":"cs.CV","submitted_at":"2026-06-28T09:40:20+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"Pointer-CAD v2 decouples planning from construction in LLM-based CAD generation by using a pointer mechanism to reference continuous parameters from a design plan, paired with new hierarchical accuracy metrics.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29239","ref_index":6,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Breaking the Rounding Trap: Securing LLMs against Quantization-Conditioned Backdoors","primary_cat":"cs.CR","submitted_at":"2026-06-28T07:06:46+00:00","verdict":"CONDITIONAL","verdict_confidence":"MODERATE","novelty_score":6.0,"formal_verification":"none","one_line_summary":"QuantGuard uses differentiable rounding control to break quantization-boundary backdoors in LLMs, reducing post-quantization attack success to clean-model levels across six models and INT8/FP4/NF4.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29215","ref_index":20,"ref_count":2,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Multi-Block Diffusion Language Models","primary_cat":"cs.LG","submitted_at":"2026-06-28T05:53:45+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"MBD-LMs raise average tokens per forward pass from 3.47 to 6.19 (and to 9.34 with DMax) via multi-block teacher forcing and optimized parallel decoding while holding or slightly improving accuracy on math and code tasks.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29194","ref_index":5,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"AI Trading's Alpha Singularity: Emergent Market Reasoning through Agent-to-Agent Self-Evolution","primary_cat":"cs.AI","submitted_at":"2026-06-28T04:41:00+00:00","verdict":"REJECT","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"Multi-agent LLM system Agora under Sealed Joint Search conditions produces +1.87 holdout Sharpe on CSI 1000 over a 91-day sealed period, exceeding the best baseline at +1.334 under favorable seed.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29184","ref_index":62,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"BaRA: Bayesian Adaptive Rank Allocation for Parameter-Efficient Fine-Tuning","primary_cat":"cs.LG","submitted_at":"2026-06-28T04:08:09+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":5.0,"formal_verification":"none","one_line_summary":"BaRA adds Bayesian adaptive rank allocation to LoRA fine-tuning by activating sparse instance-specific latent factors, with a generalization bound depending on learned joint effective rank rather than fixed maximum rank.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29155","ref_index":8,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"OASIF: An Efficient Obfuscation-Aware Self-Improving Framework for LLM-Based Assembly Code Instruction Following and Comprehension","primary_cat":"cs.SE","submitted_at":"2026-06-28T02:28:10+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"OASIF improves open-source LLMs on obfuscated assembly comprehension by 5-17 percentage points on commercial VM obfuscators via a three-phase self-evolving training pipeline.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.29066","ref_index":16,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"$x$-Prediction Flow: Efficient Continuous Decoding for Masked Diffusion Language Models","primary_cat":"cs.CL","submitted_at":"2026-06-27T19:51:35+00:00","verdict":null,"verdict_confidence":null,"novelty_score":null,"formal_verification":null,"one_line_summary":null,"context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.28998","ref_index":5,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"Reward-Free Code Alignment from Pretrained or Fine-Tuned LLM: Unpacking the Trade-offs for Code Generation","primary_cat":"cs.SE","submitted_at":"2026-06-27T16:22:22+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":4.0,"formal_verification":"none","one_line_summary":"Empirical study on five LLMs finds pretrained-to-aligned paths yield bigger gains over baseline than finetuned-to-aligned paths, though absolute accuracy remains lower for pretrained starts.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.28962","ref_index":29,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"FlipGuard: Defending Large Language Models Against Quantization-Conditioned Backdoor Attacks","primary_cat":"cs.CR","submitted_at":"2026-06-27T15:02:42+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":7.0,"formal_verification":"none","one_line_summary":"FlipGuard perturbs LLM weights prior to quantization to neutralize quantization-conditioned backdoor attacks, evaluated via the Defense Effectiveness Ratio on multiple models and quantization schemes.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null},{"citing_arxiv_id":"2606.28661","ref_index":9,"ref_count":1,"confidence":0.98,"is_internal_anchor":true,"paper_title":"When More Sampling Hurts: The Modal Ceiling and Correlation Ceiling of Test-Time Scaling","primary_cat":"cs.LG","submitted_at":"2026-06-27T00:37:33+00:00","verdict":"UNVERDICTED","verdict_confidence":"LOW","novelty_score":6.0,"formal_verification":"none","one_line_summary":"Test-time sampling improves coverage but stalls at modal and correlation ceilings for answer selection, with the effective number of samples as the practical limit.","context_count":0,"top_context_role":null,"top_context_polarity":null,"context_text":null}],"limit":100,"offset":0}}