mlsanityData to 7 Oct 2026

FrontierScience: Evaluating AI's Ability to Perform Expert-Level Scientific Tasks

arXiv 2601.21165 ↗First posted 29 Jan 2026cs.AI cs.CY cs.LGcited by 42

On the map: LLM reasoning → Knowledge- and tool-augmented reasoning

Abstract (arXiv)

We introduce FrontierScience, a benchmark evaluating expert-level scientific reasoning in frontier language models. Recent model progress has nearly saturated existing science benchmarks, which often rely on multiple-choice knowledge questions or already published information. FrontierScience addresses this gap through two complementary tracks: (1) Olympiad, consisting of international olympiad problems at the level of IPhO, IChO, and IBO, and (2) Research, consisting of PhD-level, open-ended problems representative of sub-tasks in scientific research. FrontierScience contains several hundred questions (including 160 in the open-sourced gold set) covering subfields across physics, chemistry, and biology, from quantum electrodynamics to synthetic organic chemistry. All Olympiad problems are originally produced by international Olympiad medalists and national team coaches to ensure standards of difficulty, originality, and factuality. All Research problems are research sub-tasks written and verified by PhD scientists (doctoral candidates, postdoctoral researchers, or professors). For Research, we introduce a granular rubric-based evaluation framework to assess model capabilities throughout the process of solving a research task, rather than judging only a standalone final answer.

Key facts. Cited by 42 papers in the map (data through 7 Oct 2026). Placed in Knowledge- and tool-augmented reasoning, one of 17 directions in LLM reasoning. Builds on 8 of its 11 references that are in the map. First posted on arXiv 29 Jan 2026.

Builds on

8 of 11 references.

References · 8

  1. 2020Measuring Massive Multitask Language Understandingcited by 7,768
  2. 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,748
  3. 2022Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answeringcited by 1,881
  4. 2024MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmarkcited by 1,758
  5. 2024OlympiadBench: A Challenging Benchmark for Promoting AGI with Olympiad-Level Bilingual Multimodal Scientific Problemscited by 1,099
  6. 2025PaperBench: Evaluating AI's Ability to Replicate AI Researchcited by 286
  7. 2023SciBench: Evaluating College-Level Scientific Problem-Solving Abilities of Large Language Modelscited by 246
  8. 2024LLM-Rubric: A Multidimensional, Calibrated Approach to Automated Evaluation of Natural Language Textscited by 81

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 42.

42 papers.

Sort:

By area:

  1. 2026MiroThinker-1.7 & H1: Towards Heavy-Duty Research Agents via Verificationcited by 28
  2. 2026Toward Generalist Autonomous Research via Hypothesis-Tree Refinementcited by 10
  3. 2026P1-VL: Bridging Visual Perception and Scientific Reasoning in Physics Olympiadscited by 7
  4. 2026LABBench2: An Improved Benchmark for AI Systems Performing Biology Researchcited by 6
  5. 2026Collider-Bench: Benchmarking AI Agents with Particle Physics Analysis Reproductioncited by 6
  6. 2026PRL-Bench: A Comprehensive Benchmark Evaluating LLMs' Capabilities in Frontier Physics Researchcited by 5
  7. 2026PRBench: End-to-end Paper Reproduction in Physics Researchcited by 4
  8. 2026ASI-Evolve: AI Accelerates AIcited by 4
  9. 2026Argus: Evidence Assembly for Scalable Deep Research Agentscited by 4
  10. 2026SimpleOPD: Simple Tokenizer-Agnostic On-Policy Distillation for Long-Context Reasoningcited by 4
  11. 2026MATRIX: A Multimodal Benchmark and Post-Training Framework for Materials Sciencecited by 3
  12. 2026SciPredict: Can LLMs Predict the Outcomes of Scientific Experiments in Natural Sciences?cited by 3
30 more
  1. 2026Reliable Chain-of-Thought via Prefix Consistencycited by 3
  2. 2026Economy of Minds: Emerging Multi-Agent Intelligence with Economic Interactionscited by 3
  3. 2026Agents-K1: Towards Agent-native Knowledge Orchestrationcited by 3
  4. 2026Classroom Final Exam: An Instructor-Tested Reasoning Benchmarkcited by 2
  5. 2026CITE: Anytime-Valid Statistical Inference in LLM Self-Consistencycited by 2
  6. 2026Teaching Thinking Models to Reason with Tools: A Full-Pipeline Recipe for Tool-Integrated Reasoningcited by 2
  7. 2026ProjectionBench: Evaluating Scientific Hypothesis Generation in LLMs Under Progressive Information Disclosurecited by 2
  8. 2026AgentCompass: A Unified Evaluation Infrastructure for Agent Capabilitiescited by 2
  9. 2026TopoAgent: A Self-Evolving Topological Agent for Multimodal Scientific Reasoningcited by 2
  10. 2026K-Bench: measuring model performance on real scientific agent requestscited by 2
  11. 2026Apodex 1.1: Scaling Agentic Intelligence for Complex Workcited by 2
  12. 2026Toward Evaluation Frameworks for Multi-Agent Scientific AI Systemscited by 1
  13. 2026COMPOSITE-Stemcited by 1
  14. 2026SciResearcher: Scaling Deep Research Agents for Frontier Scientific Reasoningcited by 1
  15. 2026Parametric Skillscited by 1
  16. 2026What is Missing from AI Post-Training AI: An Empirical Analysiscited by 1
  17. 2026Metacognition as Reward: Reinforcing LLM Reasoning via Knowledge and Regulation Signalscited by 0
  18. 2026Reasoning4Sciences: Bridging Reasoning Language Models to All Scientific Branchescited by 0
  19. 2026SFBench: The SciFy Scientific Feasibility Benchmarkcited by 0
  20. 2026Cross-Domain Hybrid OPD for Generalizable Search Agentscited by 0
  21. 2026MineCEraft: Evaluating Language Models as Construction Engineers in the World of Minecraftcited by 0
  22. 2026Mo' Models, Mo' Problems: How to best select model pools when designing Multi-Agent Systemscited by 0
  23. 2026Rufus-Air: An Open LLM Post-Training Recipecited by 0
  24. 2026PluginRSI: Recursive Improvement of Agent Harnesses with Reusable Pluginscited by 0
  25. 2026Beyond Token Alignment: Event Completion for Cross-Tokenizer On-Policy Distillationcited by 0
  26. 2026Long-Horizon Scaling: How Model Capabilities Shape the Returns to Computationcited by 0
  27. 2026Imprint Reader: From Weight-Update Readout to Behavioral Interventioncited by 0
  28. 2026Hermes: Learning Contextual Reasoning Unlocks Test-Time Scalingcited by 0
  29. 2026AutoDataBench: A Data-centric Testbed for Accelerating Auto Researchcited by 0
  30. 2026OpenProblemBench: Benchmarking AI on Open Problems in the Foundational Theoretical Sciencescited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery
  2. 2026Contemporary AI lacks the imagination to diverge or negate in science
  3. 2025ATLAS: A High-Difficulty, Multidisciplinary Benchmark for Frontier Scientific Reasoning
  4. 2026AIRS-Bench: a Suite of Tasks for Frontier AI Research Science Agents
  5. 2025ResearcherBench: Evaluating Deep AI Research Systems on the Frontiers of Scientific Inquiry
  6. 2025Evaluating Frontier LLMs on PhD-Level Mathematical Reasoning: A Benchmark on a Textbook in Theoretical Computer Science about Randomized Algorithms
  7. 2024LAB-Bench: Measuring Capabilities of Language Models for Biology Research
  8. 2026Scientific reasoning does not reliably translate into scientific forecasting in frontier AI