Papers

Field

Top tools here

Coming with more sources: stars history

All tools

opencode+23% downloads · 3.7M/wk+23.3%
Aider+12% downloads · 68.2k/wk+12.4%
Qwen Code13 releases this week
Claude Code6 releases this week
Gemini CLI6 releases this week

Top tools here

Open problems in Code models

Handling multi-step dependencies in code generation2 papers raise it · 1 addresses it
Real-world code changes often span multiple interdependent steps, which current benchmarks and agents struggle to model and evaluate.
- WEAVE: Weaving Real-World Code Evolution into Dependency-Structured Tasks for Benchmarking and Training Code Agentsraises · addresses
- 3DCodeVerse: Exploring 3D Worlds as Coderaises
Premise critique and abstention in code generation1 paper raises it · 1 addresses it
Models often generate executable code for invalid or unanswerable queries, leading to misleading outputs.
- PremiseSQL: Evaluating Premise Critique and Abstention in Text-to-SQL Systemsraises · addresses
Efficient verification and feedback for formal theorem proving1 paper raises it · 1 addresses it
Slow verification and timeouts hinder training and evaluation of proof-generating models.
- Learning Lean Feedback for Efficient Proof-Candidate Evaluation on GPUsraises · addresses
Multi-hop lemma retrieval for theorem proving1 paper raises it · 1 addresses it
Existing retrieval systems fail to leverage multi-hop dependencies among declarations, limiting proof construction.
- Lean Graph Search for Formal Theorem Provingraises · addresses
Context-aware infilling for multi-target code edits1 paper raises it · 1 addresses it
Causal autoregressive models struggle with edits requiring both left and right context, such as refactoring.
- Multi-Target Few-Shot Prompting with Diffusion Language Modelsraises · addresses

Named by an AI from these papers' abstracts.

Open problems

Recently funded

Market

BoxLiteSeed · Oct 19undisclosed
InfinigenceSeries B+ · Oct 12$343.6M
Qingcheng JizhiSeries A · Oct 12undisclosed
KlarentSeed · Oct 9$8M
ZenitySeries B+ · Oct 7$125M

Recently funded

Topics in this field

Active contributors · 90 days

Active contributors

Activity · 12 weeks

Coming with more sources: repos

Activity

Recent papers

All in Research →

arXiv cs.SEOct 15
WEAVE: Weaving Real-World Code Evolution into Dependency-Structured Tasks for Benchmarking and Training Code Agents
Weihao Li, Zihang Xu, Zhongqiu Wang, Songwei Yu, Wenqiang Zhu +5
Real-world software engineering rarely consists of isolated changes. Features, migrations, and systematic fixes often unfold across multiple pull requests (PRs), requiring coordinated modifications across a repository. Yet most executable repository-level benchmarks retain a single issue or PR as the task unit, while existing multi-PR settings rely on coarse aggregation and structure-agnostic evaluation. This mismatch obscures dependencies among changes and limits progress measurement and learning signals for long-horizon code agents. We introduce WEAVE, a framework that transforms verified PR histories into executable, dependency-structured tasks. WEAVE leverages repository-calibrated typed change loci to organize related PRs into bounded DAGs and instantiate each as a unified executable task. We further introduce Graph Credit, a deterministic, test-grounded metric that measures dependency-consistent partial progress and supports dense, verifiable rewards for reinforcement learning. Experiments on task construction, agent evaluation, supervised fine-tuning, and reinforcement learning show that WEAVE yields coherent executable tasks, enables more faithful progress measurement…
PDF arXiv
arXiv cs.LGOct 15
Interfaces and Modules: A Scalable Representation for Weight-Space Learning across Architectures
Michał Kozyra, Michael Bronstein
Weight-space learning across architectures requires representations that respect parameter symmetries and scale to large networks. Existing equivariant tensor encoders require architecture-specific symmetry specifications, while neuron graphs make global attention costly. We introduce a typed graph organized around shared coordinate-permutation actions. Its nodes, called interfaces, represent coordinate spaces whose bound tensor axes must be permuted jointly; operation occurrences, called modules, form typed hyperedges. A compiler derives this graph from supported PyTorch models, with exact recovery of the permutation factors defined by its constraints. The Universal Weight Encoder (UWE) combines local equivariant updates to parameter features with global attention over invariant interface summaries. The small number of interfaces makes global attention practical, and a single encoder processes different supported architectures. UWE is competitive with architecture-specific state-of-the-art methods on established benchmarks. On a new zoo of 1,200 transformers spanning 100 architectures, it outperforms the evaluated baselines at ranking networks within unseen architectures,…
PDF arXiv
arXiv cs.LGOct 15
Learning Lean Feedback for Efficient Proof-Candidate Evaluation on GPUs
Lazar Milikic, Etienne Bamas, Emmanuel Abbe, Viktor Kunčak
Training language models for formal theorem proving requires large numbers of verifier calls. Slow checks and timeouts often leave GPUs waiting and withhold rewards from potentially correct proofs. We train neural evaluators to predict Lean 4 verdicts and compiler feedback on GPUs using 4.84M labelled programs. We compare two validity classifiers with a feedback model that annotates proof attempts with predicted goal states and error messages. A two-stage evaluator that screens candidates with a classifier and scores its positive predictions with the feedback model reaches 87.8% accuracy in predicting whether Lean accepts a candidate proof, evaluated on a 68,932-program benchmark collected from public provers. The combined evaluator processes 3.5 candidates per second on one H100, 4.2x the throughput of our 64-worker Lean baseline with environment reuse. Finally, we combine our evaluator with Lean verdicts to obtain a hybrid reward system that shortens GRPO training steps by approximately 31%. After one GRPO epoch with hybrid rewards, the prover achieves nearly the same miniF2F pass@1 as with Lean-only rewards (62.2% versus 62.4%) and a pass@32 of 72.4% versus 71.3%.
PDF arXiv
arXiv cs.AIOct 15
DeciCAD: Harness-Grounded Learning of Action Policies for CAD Agents
Yufeng Peng, Yixin Zhang, Ziyi Liu, Zhikun Xu, Jiarong Yang +3
Large Language Model agents operate through closed-loop action policies whose decision space is defined by an agent harness. We investigate how this operating space shapes agent behavior and policy learning in computer-aided design (CAD), a professional domain characterized by persistent state, geometric constraints, and long-horizon procedures. We instantiate this perspective in DeciCAD, a domain-native harness that exposes semantic CAD actions, decision-oriented observations, environment feedback, and persistent state transitions. Building on this harness, we develop a learning pipeline that extracts causal decision supervision from interaction trajectories and improves the policy through supervised fine-tuning (SFT) and geometry-grounded online reinforcement learning (RL). Across four CAD benchmarks, DeciCAD achieves the best F1 and intersection over union (IoU) while reducing model-output tokens by 45--88% relative to a general-purpose coding harness. Under matched supervision, its SFT policy improves F1 from 0.6883 to 0.7437 and outperforms the coding policy; subsequent online RL further raises geometry availability from 66.5% to 73.5% and F1 to 0.7880. These results…
PDF arXiv
arXiv cs.SEOct 15
Poison Valleys: Diagnosing Path-Dependent Failure in LLM Code Repair
Veljko Skarich
When code-repair agents fail, final tests reveal that a patch is wrong but not which earlier commitment caused it or which software obligation failed. Poison Valleys traces this path in 1,844 model requests on two synthetic Python cases: PersistQueue adds atomic batch claiming to a disk-backed queue, while Dijkstra adds source-sensitive filtering to shortest-path search. We define plausible but incorrect software mechanisms, place each in a model's hypothesis, implementation plan, or both, and evaluate the generated code with an Oracle and targeted diagnostic coordinates. Moving the misconception from hypothesis to plan increased failure by 32.2% in PersistQueue and 28.9% in Dijkstra after matching the experimental mix. Different misconceptions produced distinct, model-dependent patterns of violated obligations, although several could converge on the same behavior. This creates a measured connection from upstream misconception to code mechanism and failure form. Code, data, and demo: https://poisonvalleys.org.
PDF arXiv
arXiv cs.CLOct 15
OPRD: On-Policy Reminder Distillation for Long-Horizon Coding Agents
Yuechen Luo, Zhiyi Lai, Miao Ziqi, Zihang Xu, Jingxuan Xu +7
Reliable repository-level coding requires agents not only to generate correct code but also to preserve constraints and respond to tool feedback across long interactions. However, our diagnostic analysis shows that agents often fail to act on relevant information already available in context, while final Pass/Fail outcomes provide only sparse process supervision. We introduce On-Policy Reminder Distillation (OPRD), which combines Multi-Scope Reminder supervision with outcome-based routing on reminder-free student rollouts. Successful rollouts provide autoregressive self-imitation targets, whereas failed rollouts receive Reminder-conditioned reverse-KL supervision from a teacher that re-evaluates the same student-generated prefixes. A Global Reminder converts final-test diagnostics into trajectory-level guidance, while Turn-local Reminders apply only to the next decision following the relevant environment observation. The teacher and student share a two-stage initialization from Coding SFT and Instruction-Following (IF) SFT. At inference, the student operates without Reminders. On SWE-bench Verified, SWE-bench Pro, and SWE-bench Multilingual, OPRD improves the benchmark-averaged…
PDF arXiv

Papers

Recently funded

Market

BoxLiteSeed · Oct 19undisclosed
InfinigenceSeries B+ · Oct 12$343.6M
Qingcheng JizhiSeries A · Oct 12undisclosed
KlarentSeed · Oct 9$8M
ZenitySeries B+ · Oct 7$125M

Recent papers

All in Research →

arXiv cs.SEOct 15
WEAVE: Weaving Real-World Code Evolution into Dependency-Structured Tasks for Benchmarking and Training Code Agents
Weihao Li, Zihang Xu, Zhongqiu Wang, Songwei Yu, Wenqiang Zhu +5
Real-world software engineering rarely consists of isolated changes. Features, migrations, and systematic fixes often unfold across multiple pull requests (PRs), requiring coordinated modifications across a repository. Yet most executable repository-level benchmarks retain a single issue or PR as the task unit, while existing multi-PR settings rely on coarse aggregation and structure-agnostic evaluation. This mismatch obscures dependencies among changes and limits progress measurement and learning signals for long-horizon code agents. We introduce WEAVE, a framework that transforms verified PR histories into executable, dependency-structured tasks. WEAVE leverages repository-calibrated typed change loci to organize related PRs into bounded DAGs and instantiate each as a unified executable task. We further introduce Graph Credit, a deterministic, test-grounded metric that measures dependency-consistent partial progress and supports dense, verifiable rewards for reinforcement learning. Experiments on task construction, agent evaluation, supervised fine-tuning, and reinforcement learning show that WEAVE yields coherent executable tasks, enables more faithful progress measurement…
PDF arXiv
arXiv cs.LGOct 15
Interfaces and Modules: A Scalable Representation for Weight-Space Learning across Architectures
Michał Kozyra, Michael Bronstein
Weight-space learning across architectures requires representations that respect parameter symmetries and scale to large networks. Existing equivariant tensor encoders require architecture-specific symmetry specifications, while neuron graphs make global attention costly. We introduce a typed graph organized around shared coordinate-permutation actions. Its nodes, called interfaces, represent coordinate spaces whose bound tensor axes must be permuted jointly; operation occurrences, called modules, form typed hyperedges. A compiler derives this graph from supported PyTorch models, with exact recovery of the permutation factors defined by its constraints. The Universal Weight Encoder (UWE) combines local equivariant updates to parameter features with global attention over invariant interface summaries. The small number of interfaces makes global attention practical, and a single encoder processes different supported architectures. UWE is competitive with architecture-specific state-of-the-art methods on established benchmarks. On a new zoo of 1,200 transformers spanning 100 architectures, it outperforms the evaluated baselines at ranking networks within unseen architectures,…
PDF arXiv
arXiv cs.LGOct 15
Learning Lean Feedback for Efficient Proof-Candidate Evaluation on GPUs
Lazar Milikic, Etienne Bamas, Emmanuel Abbe, Viktor Kunčak
Training language models for formal theorem proving requires large numbers of verifier calls. Slow checks and timeouts often leave GPUs waiting and withhold rewards from potentially correct proofs. We train neural evaluators to predict Lean 4 verdicts and compiler feedback on GPUs using 4.84M labelled programs. We compare two validity classifiers with a feedback model that annotates proof attempts with predicted goal states and error messages. A two-stage evaluator that screens candidates with a classifier and scores its positive predictions with the feedback model reaches 87.8% accuracy in predicting whether Lean accepts a candidate proof, evaluated on a 68,932-program benchmark collected from public provers. The combined evaluator processes 3.5 candidates per second on one H100, 4.2x the throughput of our 64-worker Lean baseline with environment reuse. Finally, we combine our evaluator with Lean verdicts to obtain a hybrid reward system that shortens GRPO training steps by approximately 31%. After one GRPO epoch with hybrid rewards, the prover achieves nearly the same miniF2F pass@1 as with Lean-only rewards (62.2% versus 62.4%) and a pass@32 of 72.4% versus 71.3%.
PDF arXiv
arXiv cs.AIOct 15
DeciCAD: Harness-Grounded Learning of Action Policies for CAD Agents
Yufeng Peng, Yixin Zhang, Ziyi Liu, Zhikun Xu, Jiarong Yang +3
Large Language Model agents operate through closed-loop action policies whose decision space is defined by an agent harness. We investigate how this operating space shapes agent behavior and policy learning in computer-aided design (CAD), a professional domain characterized by persistent state, geometric constraints, and long-horizon procedures. We instantiate this perspective in DeciCAD, a domain-native harness that exposes semantic CAD actions, decision-oriented observations, environment feedback, and persistent state transitions. Building on this harness, we develop a learning pipeline that extracts causal decision supervision from interaction trajectories and improves the policy through supervised fine-tuning (SFT) and geometry-grounded online reinforcement learning (RL). Across four CAD benchmarks, DeciCAD achieves the best F1 and intersection over union (IoU) while reducing model-output tokens by 45--88% relative to a general-purpose coding harness. Under matched supervision, its SFT policy improves F1 from 0.6883 to 0.7437 and outperforms the coding policy; subsequent online RL further raises geometry availability from 66.5% to 73.5% and F1 to 0.7880. These results…
PDF arXiv
arXiv cs.SEOct 15
Poison Valleys: Diagnosing Path-Dependent Failure in LLM Code Repair
Veljko Skarich
When code-repair agents fail, final tests reveal that a patch is wrong but not which earlier commitment caused it or which software obligation failed. Poison Valleys traces this path in 1,844 model requests on two synthetic Python cases: PersistQueue adds atomic batch claiming to a disk-backed queue, while Dijkstra adds source-sensitive filtering to shortest-path search. We define plausible but incorrect software mechanisms, place each in a model's hypothesis, implementation plan, or both, and evaluate the generated code with an Oracle and targeted diagnostic coordinates. Moving the misconception from hypothesis to plan increased failure by 32.2% in PersistQueue and 28.9% in Dijkstra after matching the experimental mix. Different misconceptions produced distinct, model-dependent patterns of violated obligations, although several could converge on the same behavior. This creates a measured connection from upstream misconception to code mechanism and failure form. Code, data, and demo: https://poisonvalleys.org.
PDF arXiv
arXiv cs.CLOct 15
OPRD: On-Policy Reminder Distillation for Long-Horizon Coding Agents
Yuechen Luo, Zhiyi Lai, Miao Ziqi, Zihang Xu, Jingxuan Xu +7
Reliable repository-level coding requires agents not only to generate correct code but also to preserve constraints and respond to tool feedback across long interactions. However, our diagnostic analysis shows that agents often fail to act on relevant information already available in context, while final Pass/Fail outcomes provide only sparse process supervision. We introduce On-Policy Reminder Distillation (OPRD), which combines Multi-Scope Reminder supervision with outcome-based routing on reminder-free student rollouts. Successful rollouts provide autoregressive self-imitation targets, whereas failed rollouts receive Reminder-conditioned reverse-KL supervision from a teacher that re-evaluates the same student-generated prefixes. A Global Reminder converts final-test diagnostics into trajectory-level guidance, while Turn-local Reminders apply only to the next decision following the relevant environment observation. The teacher and student share a two-stage initialization from Coding SFT and Instruction-Following (IF) SFT. At inference, the student operates without Reminders. On SWE-bench Verified, SWE-bench Pro, and SWE-bench Multilingual, OPRD improves the benchmark-averaged…
PDF arXiv

Activity · 12 weeks

Coming with more sources: repos

Fields · AI · updated 53m ago

Code models

Models and tools for writing, understanding, testing or repairing code, and code benchmarks.

24papers this week

All papers · 24 With code · 1

Newest Most relevant

24 papers this month

Active contributors · 90 days

Open problems in Code models

Handling multi-step dependencies in code generation2 papers raise it · 1 addresses it
Real-world code changes often span multiple interdependent steps, which current benchmarks and agents struggle to model and evaluate.
- WEAVE: Weaving Real-World Code Evolution into Dependency-Structured Tasks for Benchmarking and Training Code Agentsraises · addresses
- 3DCodeVerse: Exploring 3D Worlds as Coderaises
Premise critique and abstention in code generation1 paper raises it · 1 addresses it
Models often generate executable code for invalid or unanswerable queries, leading to misleading outputs.
- PremiseSQL: Evaluating Premise Critique and Abstention in Text-to-SQL Systemsraises · addresses
Efficient verification and feedback for formal theorem proving1 paper raises it · 1 addresses it
Slow verification and timeouts hinder training and evaluation of proof-generating models.
- Learning Lean Feedback for Efficient Proof-Candidate Evaluation on GPUsraises · addresses
Multi-hop lemma retrieval for theorem proving1 paper raises it · 1 addresses it
Existing retrieval systems fail to leverage multi-hop dependencies among declarations, limiting proof construction.
- Lean Graph Search for Formal Theorem Provingraises · addresses
Context-aware infilling for multi-target code edits1 paper raises it · 1 addresses it
Causal autoregressive models struggle with edits requiring both left and right context, such as refactoring.
- Multi-Target Few-Shot Prompting with Diffusion Language Modelsraises · addresses

Named by an AI from these papers' abstracts.

Top tools here

Coming with more sources: stars history

All tools

opencode+23% downloads · 3.7M/wk+23.3%
Aider+12% downloads · 68.2k/wk+12.4%
Qwen Code13 releases this week
Claude Code6 releases this week
Gemini CLI6 releases this week