Directory label

Reasoning benchmarks for LLMs and RL environments

A label other catalogs put on entries, not one of the domains on our map. On the map, each entry here sits in the domain its own site or paper leads with.

Hard reasoning problems built to resist memorization, such as graduate-level science questions and abstract visual puzzles.

AI benchmarks
156
Shares entries with
5 categories

Snapshot 2026-11-05. How categories are built · Where the data comes from

In short

What are the main reasoning benchmarks for LLMs?
GPQA Diamond, ARC-AGI-1, ARC-AGI-2, CritPt, AA-LCR and MuSR, among 150 confirmed below.

Only benchmarks use the Reasoning tag: the RL environment directories file their vendors under other categories. Browse all RL environments.

AI benchmarks 156

  1. GPQA Diamond
  2. ARC-AGI-1François Chollet
  3. ARC-AGI-2ARC Prize Foundation
  4. CritPtArgonne, UIUC
  5. AA-LCRArtificial Analysis
  6. MuSRUniversity of Texas at Austin
  7. LisanBench
  8. GeneBench-ProOpenAI
  9. WildBenchAllen AI
  10. Context Arena
  11. FutureEvalMetaculus
  12. ARC-AGI-3
  13. Terminal-Bench-Science
  14. Forge
  15. TRACK
  16. VeRA
  17. EnigmaEvalScale AI
  18. BBH
  19. HellaSwag
  20. WinoGrande
  21. LongBench v2THUDM
  22. Chartographer
  23. Deep20BenchPatrick Heusser, Markus Tuor, Mindalyze
  24. EgoCoT-BenchZhejiang University
  25. SimpleBench
  26. EgoMemReasonUNC Chapel Hill
  27. EstGraph
  28. LOGICAL-COMMONSENSEQA
  29. LogicScore
  30. MixRea
  31. PlanningBench
  32. SDGBiasBench
  33. SeePhys Pro
  34. StemBindNational University of Defense Technology
  35. STEMVerse
  36. TableVista
  37. TriViewBench
  38. When Seekers Are Hard to Help
  39. DROP
  40. PARSE
  41. ROSE
  42. CorpusQA 1M
  43. MRCRv2Google DeepMind
  44. Graphwalks Parents 128K
  45. MRCR 1M
  46. A2RBench
  47. Beyond Continuity
  48. ChLogic
  49. CogScaleMnemosyne
  50. CommonWhy
  51. DiscoverPhysics
  52. Enhancing Pathological VLMs with Cross-scale Reasoning
  53. LLMEval-LogicFudan NLP Lab
  54. PRISM
  55. Absurd World
  56. BilliardPhys-Bench
  57. ChaosBench-Logic v2
  58. Code-Guided Reasoning for Small Language Models
  59. CombEval
  60. Connecting the Dots
  61. CounterCount
  62. CulMind
  63. DeepWeb-BenchPeking University
  64. DeFAb
  65. Don't Judge a Book by its Cover
  66. DRInQ
  67. EnactToM
  68. From 0-Order Selection to 2-Order Judgment
  69. GENSTRAT
  70. GRACE
  71. GraphARC
  72. GRUFF: LLM Pronoun Fidelity, Reasoning, and Biases in German
  73. GSM-SEM
  74. GTEval
  75. HOLMES: Evaluating Higher-Order Logical Reasoning in LLMs
  76. HRBench
  77. Kinship Data Benchmark for Multi-hop Reasoning
  78. LEAF: A Living Benchmark for Event-Augmented Forecasting
  79. MathConstraint
  80. MINDGAMES
  81. MOV-Bench
  82. NARRA-Gym for Evaluating Interactive Narrative Agents
  83. NL-PDDL-Bench
  84. NLCO
  85. NoisyBench
  86. NoisyCausal
  87. NSMQ Riddles
  88. OmniToM
  89. ORAgentBench
  90. PCD Benchmark
  91. PIVOTS
  92. QMFOLBench
  93. ReplaySCM
  94. RMCB
  95. SciPaths: Forecasting Pathways to Scientific Discovery
  96. StressEval
  97. TeleResilienceBench
  98. Text2Opt-Bench
  99. The Complexity Ceiling Benchmark
  100. The Metanym Game
  101. TraceEval
  102. TriggerBench
  103. TS-Skill
  104. TSCognition
  105. WMW (World Models in Words)
  106. XCR-Bench
  107. XDomainBench
  108. ARC-AGI-3 Provider AdapterARC Prize Foundation
  109. STEMBenchUlam
  110. Human Behavior BenchRehearsals, Kaggle
  111. Game ArenaKaggle
  112. GIMMeta
  113. FACTS Grounding V1Google DeepMind, Jacovi et al
  114. LongMemEval
  115. GraphWalks BFS 256K–1MOpenAI
  116. Conceptual ReasoningRedwood Research
  117. Pencil Puzzle Bench
  118. MRCR v2 64K-128K
  119. MRCR v2 128K-256K
  120. MRCR v2 256K-512KOpenAI
  121. MRCR v2 512K-1M
  122. AI-NeedleKeenable.ai
  123. MLCR-AAWisedocs
  124. BullshitBench v2Peter Gostev
  125. PropensityBench
  126. NoLiMa
  127. NoLiMa-HardAdobe Research funding
  128. Long Context Reasoning (AA-LCR) Extended
  129. Long Context Reasoning (AA-LCR v1.1)
  130. FACTS Grounding v2
  131. BeliefTrackZJUNLP
  132. Can Language Model Agents be Helpful Circuit Explainers in Mechanistic Interpretability?
  133. Beyond Accuracy
  134. Block-R1
  135. ChaosBench-Logic
  136. Diagnosing and Mitigating Sycophancy and Skepticism in LLM Causal Judgment
  137. DiagnosticIQ
  138. ForeSci
  139. GeoBuildBench
  140. GR-Ben
  141. How Does Personalized Memory Shape LLM Behavior? Benchmarking Rational Preference Utilization in Personalized Assistants
  142. MIMeBench
  143. OracleProto
  144. Physics Question Scene Graph
  145. ProjectionBench
  146. Robust Asynchronous Planning via Auto-Formalization
  147. SCRuBEval
  148. The Abstraction Gap in Vision-Language Causal Reasoning
  149. VAIR
  150. XLGoBench
6 not yet confirmedListed in the catalog; no line from its own site or paper confirms it yet.
  1. Chart-FR1
  2. RULER
  3. CLUEWSC
  4. Graphwalks BFS 128K
  5. Physics-R1
  6. Why LLMs Fail at Causal Discovery and How Interventional Agents Escape

Next: pick another domain. Tags come from the RL Research daily scrape, canonicalized nightly.