Reasoning
51 milestones sit on this thread. For 35 of them it is the main thread. The others also belong to another thread, which is named under them.
2020
- MMLUMeasuring Massive Multitask Language UnderstandingMain thread: Scale and dataDataset
2021
- Codex / HumanEvalEvaluating Large Language Models Trained on CodeMain thread: Language modellingModel
2022
- Chain of thoughtChain-of-Thought Prompting Elicits Reasoning in Large Language ModelsTechnique
- AlphaCodeCompetition-Level Code Generation with AlphaCodeMain thread: Language modellingModel
- Self-consistencySelf-Consistency Improves Chain of Thought ReasoningTechnique
- STaRSTaR: Bootstrapping Reasoning With ReasoningMain thread: AlignmentTraining method
- PaLMPaLM: Scaling Language Modeling with PathwaysMain thread: Language modellingModel
- Zero-shot CoTLarge Language Models are Zero-Shot ReasonersTechnique
- Emergent abilitiesEmergent Abilities of Large Language ModelsMain thread: Scale and dataResearch paper
- ReActReAct: Synergizing Reasoning and Acting in Language ModelsMain thread: Representation and retrievalTechnique
2023
- ReflexionReflexion: Language Agents with Verbal Reinforcement LearningResearch paper
- Emergence mirageAre Emergent Abilities of Large Language Models a Mirage?Main thread: Scale and dataResearch paper
- Tree of thoughtsTree of Thoughts: Deliberate Problem Solving with Large Language ModelsTechnique
- Process supervisionLet's Verify Step by StepMain thread: Scale and dataResearch paper
- SWE-benchSWE-bench: Can Language Models Resolve Real-World GitHub Issues?Dataset
2024
- GRPODeepSeekMath: Pushing the Limits of Mathematical ReasoningMain thread: AlignmentTraining method
- Llama 3.1 405BThe Llama 3 Herd of ModelsMain thread: Scale and dataModel
- Test-time computeScaling LLM Test-Time Compute OptimallyMain thread: Scale and dataResearch paper
- o1OpenAI o1Main thread: AlignmentModel
- o3 announcedOpenAI o3 announcementModel
2025
- DeepSeek-R1DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement LearningMain thread: Language modellingModel
- Claude 3.7Claude 3.7 SonnetMain thread: AlignmentModel
- Gemini 2.5Gemini 2.5Model
- o3 releasedOpenAI o3Model
- o4-miniOpenAI o4-miniMain thread: Efficiency and infrastructureModel
- Qwen3Qwen3Model
- Claude 4Claude 4 (Opus 4 and Sonnet 4)Model
- IMO goldIMO gold-medal performanceBenchmark
- gpt-ossgpt-oss-120b and gpt-oss-20bModel
- GPT-5GPT-5Model
- GLM-4.5GLM-4.5Model
- Claude Sonnet 4.5Claude Sonnet 4.5Model
- IMO-ProofBenchTowards Robust Mathematical Reasoning (IMO-ProofBench)Research paper
- GPT-5.1GPT-5.1Model
- Gemini 3Gemini 3Model
- Claude Opus 4.5Claude Opus 4.5Model
- GPT-5.2GPT-5.2Model
2026
- Claude Opus 4.6Claude Opus 4.6Model
- Gemini 3 Deep ThinkGemini 3 Deep ThinkModel
- Qwen3.5Qwen3.5Model
- Gemini 3.1Gemini 3.1 ProModel
- GPT-5.4GPT-5.4Model
- Muse SparkMuse SparkModel
- Claude Opus 4.7Claude Opus 4.7Model
- GPT-5.5GPT-5.5Model
- Claude Opus 4.8Claude Opus 4.8Model
- Claude Sonnet 5Claude Sonnet 5Model
- GLM-5.2GLM-5.2Model
- GPT-5.6 SolGPT-5.6 SolModel
- Claude Opus 5Claude Opus 5Model
- Claude Opus 5.5Claude Opus 5.5Model