AutoLab benchmark shows frontier models mostly fail at sustained iterative optimization due to premature termination, with persistence as the key success factor.
hub Canonical reference
R2E-gym: Procedural environments and hybrid verifiers for scaling open-weights SWE agents
Canonical reference. 83% of citing Pith papers cite this work as background.
25
Pith papers citing it
Background
83% of classified citations
abstract
Improving open-source models on real-world SWE tasks (solving GITHUB issues) faces two key challenges: 1) scalable curation of execution environments to train these models, and, 2) optimal scaling of test-time compute. We introduce AgentGym, the largest procedurally-curated executable gym environment for training real-world SWE-agents, consisting of more than 8.7K tasks. AgentGym is powered by two main contributions: 1) SYNGEN: a synthetic data curation recipe that enables scalable curation of executable environments using test-generation and back-translation directly from commits, thereby reducing reliance on human-written issues or unit tests. We show that this enables more scalable training leading to pass@1 performance of 34.4% on SWE-Bench Verified benchmark with our 32B model. 2) Hybrid Test-time Scaling: we provide an in-depth analysis of two test-time scaling axes; execution-based and execution-free verifiers, demonstrating that they exhibit complementary strengths and limitations. Test-based verifiers suffer from low distinguishability, while execution-free verifiers are biased and often rely on stylistic features. Surprisingly, we find that while each approach individually saturates around 42-43%, significantly higher gains can be obtained by leveraging their complementary strengths. Overall, our approach achieves 51% on the SWE-Bench Verified benchmark, reflecting a new state-of-the-art for open-weight SWE-agents and for the first time showing competitive performance with proprietary models such as o1, o1-preview and sonnet-3.5-v2 (with tools). We will open-source our environments, models, and agent trajectories.
hub tools
citation-role summary
background 4
dataset 2
citation-polarity summary
representative citing papers
Dockerless uses agentic repository exploration to verify patches without execution, enabling SFT and RL training of coding agents that reach 62.0/50.0/35.2% resolve rates on SWE-bench Verified/Multilingual/Pro while matching environment-based results.
StaminaBench evaluates coding agents over 100 procedurally generated change requests to a REST API, finding that tested models fail within 5-6 turns without feedback but improve up to 12x with test feedback and good harnesses.
CUA-Gym generates 32,112 verified RLVR tuples across 110 mock environments, enabling trained models to reach 62.1% and 72.6% on OSWorld-Verified while transferring to WebArena.
BootstrapAgent distills repository bootstrapping heuristics into a persistent .bootstrap contract via multi-agent evidence extraction, Docker verification, and trace-driven repair, reporting 92.9% success and efficiency gains on three benchmarks.
FrontierSmith automates synthesis of open-ended coding problems from closed-ended seeds and shows measurable gains on two open-ended LLM coding benchmarks.
AgentLens reveals 10.7% of passing SWE-agent trajectories exhibit Lucky Pass behaviors and introduces a process-level evaluation framework with a new annotated dataset of 1,815 trajectories.
EO-Gym supplies an executable multimodal environment and 9k-trajectory benchmark that turns Earth Observation into a tool-using, multi-step reasoning task, revealing that current VLMs struggle on temporal and cross-sensor workflows while fine-tuning lifts Pass@3 from 0.49 to 0.74.
The first empirical study of test overfitting shows that auto-generated tests from issues can lead to code that passes observed tests but misses important cases or breaks functionality in SWE-bench issue resolution.
An agentic code reviewer that explores repositories to judge and diagnose AI-generated pull requests improves resolve rates from 27.5% to 56.9% and outperforms single-turn review baselines.
Nexus Sampling combines iterative Nexus scoring with weighted reservoir sampling to evict KV cache tokens, theoretically dominating top-K selection in retaining subtly important tokens and empirically matching dense attention within 1% at 80% eviction on LongBench.
PROPEL amortizes solver evaluation with a trained activation probe to optimize task generators toward a target solve rate, raising the share of learnable tasks from ~10% to ~20% in coding and SWE experiments.
Trajectories from weaker agents outperform stronger ones for training terminal agents due to environment-grounded supervision that exposes inspect-act-verify behaviors.
P2T distills reference patches into a latent process graph and uses it to select shortest effective trajectory segments from teacher rollouts, yielding up to 10.8 point Pass@1 gains on SWE-bench Verified with 15% lower inference cost using only 1.8k instances.
Prefix Sampling replays self-generated trajectory prefixes to control rollout pass rates near 50% in binary-reward RL, delivering wall-clock speedups and modest performance gains on SWE-bench Verified and AIME tasks.
A single LLM agent that injects and repairs its own bugs in real repositories improves SWE-bench Verified and Pro by +10.4 and +7.8 points, outperforming a human-data RL baseline.
TestPrune minimizes regression test suites to improve bug reproduction and patch validation in LLM-based agentic repair pipelines, delivering 6-13% relative gains on SWE-Bench benchmarks at low API cost.
Injecting 1% targeted synthetic data into GPT-2's pre-training substantially improves performance on 8 of 9 failing BLiMP grammatical paradigms, indicating data scarcity causes formal linguistic failures.
This survey categorizes agentic environments for LLMs by eight attributes and domains, introduces symbolic and neural synthesis paradigms with evaluation, and outlines four agent evolution pathways plus three environment evolution paradigms.
STITCH trains superior agentic coding and reasoning LLMs by using fewer high-quality trajectories filtered to keep only critical decision tokens, delivering up to 63% relative gains on SWE-bench Verified.
Survey framing LLM agents as model-plus-harness systems, decomposing harness responsibilities, mapping them to tasks, and highlighting open challenges in evaluation, safety, and co-evolution.
Libra optimizes GPU allocation across rollout and training in agentic RL via an elastic hybrid pool and C-MLFQ scheduler based on tool-return causal signals, claiming up to 3.0x throughput and 2.5x faster reward convergence on 48 A800 GPUs.
EVT improves the RMT backbone by using Euclidean-distance attention decay and 1D token grouping, achieving 86.6% top-1 on ImageNet-1K at 384×384 resolution.
A survey consolidating benchmarks, agent frameworks, real-world applications, and protocols for LLM-based autonomous agents into a proposed taxonomy with recommendations for future research.
citing papers explorer
-
AutoLab: Can Frontier Models Solve Long-Horizon Auto Research and Engineering Tasks?
AutoLab benchmark shows frontier models mostly fail at sustained iterative optimization due to premature termination, with persistence as the key success factor.
-
Dockerless: Environment-Free Program Verifier for Coding Agents
Dockerless uses agentic repository exploration to verify patches without execution, enabling SFT and RL training of coding agents that reach 62.0/50.0/35.2% resolve rates on SWE-bench Verified/Multilingual/Pro while matching environment-based results.
-
StaminaBench: Stress-Testing Coding Agents over 100 Interaction Turns
StaminaBench evaluates coding agents over 100 procedurally generated change requests to a REST API, finding that tested models fail within 5-6 turns without feedback but improve up to 12x with test feedback and good harnesses.
-
CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agents
CUA-Gym generates 32,112 verified RLVR tuples across 110 mock environments, enabling trained models to reach 62.1% and 72.6% on OSWorld-Verified while transferring to WebArena.
-
BootstrapAgent: Distilling Repository Setup into Reusable Agent Knowledge
BootstrapAgent distills repository bootstrapping heuristics into a persistent .bootstrap contract via multi-agent evidence extraction, Docker verification, and trace-driven repair, reporting 92.9% success and efficiency gains on three benchmarks.
-
FrontierSmith: Synthesizing Open-Ended Coding Problems at Scale
FrontierSmith automates synthesis of open-ended coding problems from closed-ended seeds and shows measurable gains on two open-ended LLM coding benchmarks.
-
AgentLens: Revealing The Lucky Pass Problem in SWE-Agent Evaluation
AgentLens reveals 10.7% of passing SWE-agent trajectories exhibit Lucky Pass behaviors and introduces a process-level evaluation framework with a new annotated dataset of 1,815 trajectories.
-
EO-Gym: A Multimodal, Interactive Environment for Earth Observation Agents
EO-Gym supplies an executable multimodal environment and 9k-trajectory benchmark that turns Earth Observation into a tool-using, multi-step reasoning task, revealing that current VLMs struggle on temporal and cross-sensor workflows while fine-tuning lifts Pass@3 from 0.49 to 0.74.
-
Investigating Test Overfitting on SWE-bench
The first empirical study of test overfitting shows that auto-generated tests from issues can lead to code that passes observed tests but misses important cases or breaks functionality in SWE-bench issue resolution.
-
SWE-Review: Closing the Loop on Issue Resolution with Agentic Code Review
An agentic code reviewer that explores repositories to judge and diagnose AI-generated pull requests improves resolve rates from 27.5% to 56.9% and outperforms single-turn review baselines.
-
Forget Without Compromise: Nexus Sampling for Streaming KV-Cache Eviction Under Fixed Budgets
Nexus Sampling combines iterative Nexus scoring with weighted reservoir sampling to evict KV cache tokens, theoretically dominating top-K selection in retaining subtly important tokens and empirically matching dense attention within 1% at 80% eviction on LongBench.
-
Breaking the Solver Bottleneck: Training Task Generators at the Learnable Frontier
PROPEL amortizes solver evaluation with a trained activation probe to optimize task generators toward a target solve rate, raising the share of learnable tasks from ~10% to ~20% in coding and SWE experiments.
-
What Makes Interaction Trajectories Effective for Training Terminal Agents?
Trajectories from weaker agents outperform stronger ones for training terminal agents due to environment-grounded supervision that exposes inspect-act-verify behaviors.
-
From Patches to Trajectories: Privileged Process Supervision for Software-Engineering Agents
P2T distills reference patches into a latent process graph and uses it to select shortest effective trajectory segments from teacher rollouts, yielding up to 10.8 point Pass@1 gains on SWE-bench Verified with 15% lower inference cost using only 1.8k instances.
-
Rollout Pass-Rate Control: Steering Binary-Reward RL Toward Its Most Informative Regime
Prefix Sampling replays self-generated trajectory prefixes to control rollout pass rates near 50% in binary-reward RL, delivering wall-clock speedups and modest performance gains on SWE-bench Verified and AIME tasks.
-
Toward Training Superintelligent Software Agents through Self-Play SWE-RL
A single LLM agent that injects and repairs its own bugs in real repositories improves SWE-bench Verified and Pro by +10.4 and +7.8 points, outperforming a human-data RL baseline.
-
Can Old Tests Do New Tricks for Resolving SWE Issues?
TestPrune minimizes regression test suites to improve bug reproduction and patch validation in LLM-based agentic repair pipelines, delivering 6-13% relative gains on SWE-Bench benchmarks at low API cost.
-
LiteResearcher: A Scalable Agentic RL Training Framework for Deep Research Agent
Injecting 1% targeted synthetic data into GPT-2's pre-training substantially improves performance on 8 of 9 failing BLiMP grammatical paradigms, indicating data scarcity causes formal linguistic failures.
-
Agentic Environment Engineering for Large Language Models: A Survey of Environment Modeling, Synthesis, Evaluation, and Application
This survey categorizes agentic environments for LLMs by eight attributes and domains, introduces symbolic and neural synthesis paradigms with evaluation, and outlines four agent evolution pathways plus three environment evolution paradigms.
-
Yet Even Less Is Even Better For Agentic, Reasoning, and Coding LLMs
STITCH trains superior agentic coding and reasoning LLMs by using fewer high-quality trajectories filtered to keep only critical decision tokens, delivering up to 63% relative gains on SWE-bench Verified.
-
From Question Answering to Task Completion: A Survey on Agent System and Harness Design
Survey framing LLM agents as model-plus-harness systems, decomposing harness responsibilities, mapping them to tasks, and highlighting open challenges in evaluation, safety, and co-evolution.
-
Libra: Efficient Resource Management for Agentic RL Post-Training
Libra optimizes GPU allocation across rollout and training in agentic RL via an elastic hybrid pool and C-MLFQ scheduler based on tool-return causal signals, claiming up to 3.0x throughput and 2.5x faster reward convergence on 48 A800 GPUs.
-
ClawEnvKit: Automatic Environment Generation for Claw-Like Agents
EVT improves the RMT backbone by using Euclidean-distance attention decay and 1D token grouping, achieving 86.6% top-1 on ImageNet-1K at 384×384 resolution.
-
From LLM Reasoning to Autonomous AI Agents: A Comprehensive Review
A survey consolidating benchmarks, agent frameworks, real-world applications, and protocols for LLM-based autonomous agents into a proposed taxonomy with recommendations for future research.
-
A Survey of Reinforcement Learning for Large Reasoning Models
A survey compiling RL methods, challenges, data resources, and applications for enhancing reasoning in large language models and large reasoning models since DeepSeek-R1.