IdleSpec improves LLM agent accuracy by generating and aggregating speculative plans during idle time between tool calls and observations using complementary drafting strategies.
hub Mixed citations
MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering
Mixed citation behavior. Most common role is background (67%).
53
Pith papers citing it
Background
67% of classified citations
abstract
We introduce MLE-bench, a benchmark for measuring how well AI agents perform at machine learning engineering. To this end, we curate 75 ML engineering-related competitions from Kaggle, creating a diverse set of challenging tasks that test real-world ML engineering skills such as training models, preparing datasets, and running experiments. We establish human baselines for each competition using Kaggle's publicly available leaderboards. We use open-source agent scaffolds to evaluate several frontier language models on our benchmark, finding that the best-performing setup--OpenAI's o1-preview with AIDE scaffolding--achieves at least the level of a Kaggle bronze medal in 16.9% of competitions. In addition to our main results, we investigate various forms of resource scaling for AI agents and the impact of contamination from pre-training. We open-source our benchmark code (github.com/openai/mle-bench/) to facilitate future research in understanding the ML engineering capabilities of AI agents.
hub tools
citation-role summary
background 10
dataset 4
other 1
citation-polarity summary
representative citing papers
Evolutionary coding agents achieve most benchmark gains through a small subset of edit types and by cycling previously deleted code lines rather than developing new algorithmic structures.
DecisionBench supplies a fixed task suite, model pool, delegation interface, and multi-axis metrics to evaluate emergent delegation, showing similar quality across awareness conditions but 15-31 point headroom under perfect delegation.
WebGameBench is a new benchmark that evaluates coding agents on building browser-native games from frozen specifications, with runtime browser evaluation showing best agents reach 76.9% usable rate but only 20.2% excellent rate.
BioXArena benchmarks LLM agents on generating end-to-end ML pipelines for 76 multi-modal biomedical tasks, with MLEvolve plus Gemini-3.1-Pro scoring highest at 0.666.
SMCEvolve applies Sequential Monte Carlo sampling to LLM program search with adaptive resampling, mutation mixtures, and convergence control, delivering finite-sample complexity bounds and benchmark gains over prior systems.
FrontierSmith automates synthesis of open-ended coding problems from closed-ended seeds and shows measurable gains on two open-ended LLM coding benchmarks.
Collider-Bench is a new benchmark showing that current LLM agents cannot reliably reproduce LHC analyses at the level of a physicist-in-the-loop.
BenchJack audits 10 AI agent benchmarks, synthesizes exploits achieving near-perfect scores without task completion, surfaces 219 flaws, and reduces hackable-task ratios to under 10% on four benchmarks via iterative patching.
KompeteAI accelerates AutoML pipeline evaluation 6.9 times and beats prior systems by 3% on MLE-Bench through candidate merging, external RAG, and predictive early scoring.
Frontier models demonstrate in-context scheming by strategically deceiving in multiple agentic evaluations to achieve given goals.
Enforcing role separation in agent teams reveals that prompt-only setups hide coordination failures, with verifiers approving 49% of failing work and teams sometimes harming performance when solo agents already succeed.
AcademiClaw is a new benchmark of 80 student-sourced academic tasks where the best frontier AI agents achieve only a 55% pass rate.
LLMs predict outcomes of real scientific experiments at 14-26% accuracy, comparable to human experts, but lack calibration on prediction reliability while humans demonstrate strong calibration.
FrontierFinance benchmark shows human financial experts outperform state-of-the-art LLMs by achieving higher scores and more client-ready outputs on realistic long-horizon tasks.
SERL selectively reweights learning using task success and environment feedback to reach 90.0% success on ALFWorld and 80.1% on WebShop, outperforming RL and distillation baselines.
ResearchArena shows that agent-generated papers fail top-tier acceptance standards primarily due to fabricated results, underpowered experiments, and plan-execution mismatches that vary sharply by agent.
DiagEval applies trajectory-conditioned diagnostic probes to recover 45.6-62.1% of misattributed failures in GUI-agent software evaluation, raising accuracy from 69.9% to 78.3% on WebDevJudge-Unit and 65.0% to 81.6% on RealDevBench.
MLReplicate benchmark evaluates six autonomous systems on 45 manuscripts from ICML 2025 papers, finding that automated reviews accept flawed outputs with fabricated claims while human review exposes methodological failures, and that the cheapest system outperforms the most expensive by a wide margin
DataMaster deploys an AI agent to autonomously engineer data via tree search over external sources, shared candidate pools, and memory of past outcomes, yielding 32% higher medal rates on MLE-Bench Lite and a small GPQA gain over the base instruct model.
Gome reaches 35.1% any-medal rate on MLE-Bench by mapping reasoning to gradient-based updates, outperforming tree search once models are sufficiently capable.
xKG is a paper-centric knowledge base that extracts code and insights to improve LLM agent performance on AI research replication by 10.9% on PaperBench.
MachineLearningLM uses continued pretraining on SCM-synthesized ML tasks with random-forest distillation to give LLMs robust many-shot in-context learning on tabular classification, reaching random-forest accuracy levels while preserving general chat performance.
RExBench is a new benchmark showing that LLM coding agents fail to autonomously implement most realistic research extensions to prior AI papers.
citing papers explorer
-
What Do Evolutionary Coding Agents Evolve?
Evolutionary coding agents achieve most benchmark gains through a small subset of edit types and by cycling previously deleted code lines rather than developing new algorithmic structures.