mlsanityData to 14 Oct 2026

ARC-AGI-2: A New Challenge for Frontier AI Reasoning Systems

arXiv 2505.11831 ↗First posted 17 May 2025cs.AIcited by 190

On the map: LLM reasoning → Multimodal reasoning

Abstract (arXiv)

The Abstraction and Reasoning Corpus for Artificial General Intelligence (ARC-AGI), introduced in 2019, established a challenging benchmark for evaluating the general fluid intelligence of artificial systems via a set of unique, novel tasks only requiring minimal prior knowledge. While ARC-AGI has spurred significant research activity over the past five years, recent AI progress calls for benchmarks capable of finer-grained evaluation at higher levels of cognitive complexity. We introduce ARC-AGI-2, an upgraded version of the benchmark. ARC-AGI-2 preserves the input-output pair task format of its predecessor, ensuring continuity for researchers. It incorporates a newly curated and expanded set of tasks specifically designed to provide a more granular signal to assess abstract reasoning and problem-solving abilities at higher levels of fluid intelligence. To contextualize the difficulty and characteristics of ARC-AGI-2, we present extensive results from human testing, providing a robust baseline that highlights the benchmark's accessibility to human intelligence, yet difficulty for current AI systems. ARC-AGI-2 aims to serve as a next-generation tool for rigorously measuring progress towards more general and human-like AI capabilities.

Key facts. Cited by 190 papers in the map (data through 14 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 3 of its 9 references that are in the map. First posted on arXiv 17 May 2025.

Builds on

3 of 9 references.

Closest ancestors · 1

Same direction, newest first

  1. 2024ARC Prize 2024: Technical Report5 Dec

Other references · 2

  1. 2021Fast and flexible: Human program induction in abstract reasoning taskscited by 33
  2. 2024H-ARC: A Robust Estimate of Human Performance on the Abstraction and Reasoning Corpus Benchmarkcited by 33

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 57; 2026: 133.

190 papers.

Sort:

By area:

  1. 2025Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoningcited by 559
  2. 2025The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexitycited by 302
  3. 2025Hierarchical Reasoning Modelcited by 177
  4. 2025REASONING GYM: Reasoning Environments for Reinforcement Learning with Verifiable Rewardscited by 130
  5. 2025Revisiting Reinforcement Learning for LLM Reasoning from A Cross-Domain Perspectivecited by 76
  6. 2026AdaEvolve: Adaptive LLM Driven Zeroth-Order Optimizationcited by 60
  7. 2026EvoX: Meta-Evolution for Automated Discoverycited by 60
  8. 2026Agent-World: Scaling Real-World Environment Synthesis for Evolving General Agent Intelligencecited by 54
  9. 2025Less is More: Recursive Reasoning with Tiny Networkscited by 45
  10. 2025Thinking with Video: Video Generation as a Promising Multimodal Reasoning Paradigmcited by 42
  11. 2026OfficeQA Pro: An Enterprise Benchmark for End-to-End Grounded Reasoningcited by 39
  12. 2025Recursive Self-Aggregation Unlocks Deep Thinking in Large Language Modelscited by 35
38 more
  1. 2026When AI Benchmarks Plateau: A Systematic Study of Benchmark Saturationcited by 33
  2. 2026ClawsBench: Evaluating Capability and Safety of LLM Productivity Agents in Simulated Workspacescited by 33
  3. 2025PRBench: Large-Scale Expert Rubrics for Evaluating High-Stakes Professional Reasoningcited by 32
  4. 2026Reinforcing Chain-of-Thought Reasoning with Self-Evolving Rubricscited by 26
  5. 2025Assessing Adaptive World Models in Machines with Novel Gamescited by 20
  6. 2026Generative Recursive Reasoningcited by 20
  7. 2026Probabilistic Tiny Recursive Modelcited by 19
  8. 2025BOTS: A Unified Framework for Bayesian Online Task Selection in LLM Reinforcement Finetuningcited by 14
  9. 2025Bottom-up Domain-specific Superintelligence: A Reliable Knowledge Graph is What We Needcited by 13
  10. 2026VLMs Need Words: Vision Language Models Ignore Visual Detail In Favor of Semantic Anchorscited by 13
  11. 2026Rethinking Reward Supervision: Rubric-Conditioned Self-Distillationcited by 13
  12. 2026Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Modelscited by 12
  13. 2025Universal Reasoning Modelcited by 11
  14. 2026Jet-RL: Enabling On-Policy FP8 Reinforcement Learning with Unified Training and Rollout Precision Flowcited by 11
  15. 2026F-GRPO: Don't Let Your Policy Learn the Obvious and Forget the Rarecited by 11
  16. 2026Harnessing Agentic Evolutioncited by 11
  17. 2025Do AI Models Perform Human-like Abstract Reasoning Across Modalities?cited by 10
  18. 2025ARC-GEN: A Mimetic Procedural Benchmark Generator for the Abstraction and Reasoning Corpuscited by 10
  19. 2026PRO-LONG: Programmatic Memory Enables Long-Horizon Reasoningcited by 10
  20. 2025Comprehension Without Competence: Architectural Limits of LLMs in Symbolic Computation and Reasoningcited by 9
  21. 2025GVGAI-LLM: Evaluating Large Language Model Agents with Infinite Gamescited by 9
  22. 202511Plus-Bench: Demystifying Multimodal LLM Spatial Reasoning with Cognitive-Inspired Analysiscited by 9
  23. 2025Tiny Recursive Models on ARC-AGI-1: Inductive Biases, Identity Conditioning, and Test-Time Computecited by 9
  24. 2025Position: Reasoning After Perception Means Reasoning Without Visioncited by 8
  25. 2026Squeeze Evolve: Unified Multi-Model Orchestration for Verifier-Free Evolutioncited by 8
  26. 2026Loop the Loopies!cited by 8
  27. 2025PRELUDE: A Benchmark Designed to Require Global Comprehension and Reasoning over Long Contextscited by 7
  28. 2025VeriEquivBench: An Equivalence Score for Ground-Truth-Free Evaluation of Formally Verifiable Codecited by 7
  29. 2025HardcoreLogic: Challenging Large Reasoning Models with Long-tail Logic Puzzle Gamescited by 7
  30. 2026Toward Cognitive Supersensing in Multimodal Large Language Modelcited by 7
  31. 2025Truly Assessing Fluid Intelligence of Large Language Models through Dynamic Reasoning Evaluationcited by 6
  32. 2025MORSE-500: A Programmatically Controllable Video Benchmark to Stress-Test Multimodal Reasoningcited by 6
  33. 2026Learning from Synthetic Data Improves Multi-hop Reasoningcited by 6
  34. 2026Measuring Progress Toward AGI: A Cognitive Frameworkcited by 6
  35. 2026From Chatbot to Digital Colleague: The Paradigm Shift Toward Persistent Autonomous AIcited by 6
  36. 2026Ling and Ring 2.6 Technical Report: Efficient and Instant Agentic Intelligence at Trillion-Parameter Scalecited by 6
  37. 2025Neural Cellular Automata: From Cells to Pixelscited by 5
  38. 2025Reasoning Core: A Scalable RL Environment for LLM Symbolic Reasoningcited by 5

50 most cited of 190 All 190 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026ARC-AGI-3: A New Challenge for Frontier Agentic Intelligence
  2. 2025ARC-NCA: Towards Developmental Solutions to the Abstraction and Reasoning Corpus
  3. 2023Unraveling the ARC Puzzle: Mimicking Human Solutions with Object-Centric Decision Transformer
  4. 2025Product of Experts with LLMs: Boosting Performance on ARC Is a Matter of Perspective
  5. 2024Neural networks for abstraction and reasoning: Towards broad generalization in machines
  6. 2026ARC-TGI: Human-Validated Task Generators with Reasoning Chain Templates for ARC-AGI
  7. 2023Abstract Visual Reasoning Enabled by Language
  8. 2025Don't throw the baby out with the bathwater: How and why deep learning for ARC