6arXiv cs.CL (Computation and Language)·
RSIBench-Data: Benchmark for LLM Agents as Data-Centric Researchers in Recursive Self-Improvement
RSIBench-Data is a new benchmark that isolates the data-centric research capabilities of LLM agents within a fixed post-training stack, testing whether agents can autonomously diagnose capability gaps and revise training-data strategies to improve a target model. Four frontier agents are evaluated across six benchmarks spanning software engineering, math, and scientific QA. Agents improve on their first valid attempt in 58% of settings, but 78% of runs that continue after a peak score end lower, indicating inconsistent self-improvement. The benchmark is open-sourced and provides a controlled testbed for measuring progress toward recursive self-improvement.
Related guides (3)
Related events (8)
AARRI-Bench evaluates frontier LLMs and agents on granular research-intern-level tasks
Researchers introduce AARR (Act As a Real Researcher), a new benchmark series targeting whether AI agents can emulate the professionalism, thoroughness, and nuanced judgment of human researchers in granular research scenarios—not just macro-level task execution. The first benchmark, AARRI-Bench, tests frontier models and agentic harnesses, finding that even the best configuration (Mini-SWE-Agent with Claude Opus 4.7) achieves only 68.3% success, frequently missing subtle but critical details obvious to human researchers. The work argues that closing the gap requires deeper modeling of research behavior rather than more complex scaffolding.
RRBench: evaluating open-weight LLMs on longitudinal data preparation tasks in governance-restricted settings
Researchers from UCL introduce RRBench, an open-source benchmark and evaluation framework for assessing open-weight LLMs on data preparation tasks in longitudinal population studies, where cloud-based models are prohibited by data governance rules. The benchmark uses ground-truth cleaning scripts from six sweeps of a British cohort study, covering tasks like category harmonization and multi-wave merging, with automated evaluation of LLM-generated R code. Current 31–35B parameter open-weight models achieve up to 87.9% average task completion on consumer-grade hardware, suggesting a viable path for AI-assisted data preparation in privacy-sensitive research contexts.
RevengeBench: Benchmark for Reconstructing Agent Decision Programs from Behavioral Observations
RevengeBench is a new benchmark of 75 LLM-generated, Elo-calibrated policies across five game environments that tests whether a learner can reconstruct a hidden agent's decision program as executable code from behavioral traces alone. The benchmark draws from CodeClash tournament trajectories and allows the learner to design controlled behavioral probes (custom opponent policies) to elicit informative behavior before submitting an executable hypothesis. Evaluated across twelve frontier LLMs, recovery quality ranges from 34 to 72% of initial action-distance closed, with reconstructed policies providing measurable competitive advantage especially for weaker models. The work frames policy reconstruction as a tractable inverse problem in code-space, with implications for opponent modeling and policy interpretability.
Benchmark Agent: Autonomous system for end-to-end benchmark construction
Researchers introduce Benchmark Agent, a fully autonomous agentic system that orchestrates the complete benchmark construction pipeline — from query analysis and subtask design to data annotation and quality control. The system was used to produce 15 benchmarks spanning text understanding, multimodal understanding, and domain-specific reasoning, with evaluation via human judges, LLM-as-a-judge, and consistency checks. The work addresses two persistent problems in the field: the labor intensity of benchmark creation and rapid performance saturation after release. Code and a demo will be publicly released.
Double Ratchet: Co-evolving evaluation metrics and skills for self-improving LLM agents
A new arXiv preprint introduces Double Ratchet, a system that co-evolves both evaluation metrics and agent skills in settings where no reliable automatic verifier exists. The metric loop uses evolutionary search over small drawback detectors anchored to a small reference set, while the skill loop uses a lifecycle-managed approach; together they retain 88–110% of the performance lift achievable with ground-truth metrics across code generation (MBPP+), text-to-SQL (Spider 2.0-Snow), and report generation tasks. The paper also addresses safety, showing that anchor discipline and outer audits can catch and repair cases where evolved skills game the rubric. This work directly addresses a core bottleneck in self-improving agent systems: the chicken-and-egg problem of needing a reliable evaluator to improve.
SkillGenBench: Benchmarking Skill Generation Pipelines for LLM Agents
SkillGenBench is a new benchmark designed to evaluate the ability of LLM agents to generate correct, reusable, and executable skills from raw repositories and documents, rather than merely using pre-provided skills. It covers two generation regimes (task-conditioned and task-agnostic) and two procedural sources (repository-grounded and document-grounded), with standardized execution-based evaluation protocols. Experiments across multiple skill-generation methods reveal substantial performance variation and distinct failure modes depending on source type. The benchmark aims to establish skill generation as an independent research problem within agent systems.
RePro: Retrospective Progress-Aware Self-Refinement for LLM Agent Training
Researchers introduce RePro (Retrospective Progress-Aware Training), a framework addressing the gap between step-wise RL optimization and metacognitive task-progress awareness in LLM agents. The approach uses a forward-then-reflect rollout paradigm where agents execute actions online and then retrospectively assess step-wise progress given the completed trajectory and known outcome. Evaluated on WebShop, ALFWorld, and Sokoban, RePro achieves up to 12% absolute success rate gains over baseline Qwen-family models without requiring continuous external supervision.
MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering
OpenAI introduces MLE-bench, a benchmark designed to measure AI agent performance on machine learning engineering tasks. The benchmark draws from Kaggle competitions to evaluate agents on realistic ML engineering workflows. Initial results show that current agents, including those powered by o1-preview, achieve competitive performance on a subset of tasks but fall well short of top human competitors. The benchmark is intended to track progress in agentic ML capabilities over time.