Multimodal Chain-of-Thought Reasoning: A Comprehensive Survey.

[code]

Recent Advances in Large Language Model Benchmarks Against Data Contamination: From Static to Dynamic Evaluation.

[code]

Attention Heads of Large Language Models: A Survey.

[code]

Internal Consistency and Self-Feedback in Large Language Models: A Survey.

[code]

Puzzle Solving using Reasoning of Large Language Models: A Survey.

[code]

Large Language Models for Mathematical Reasoning: Progresses and Challenges.

Towards Reasoning in Large Language Models: A Survey.

[code]

Reasoning with Language Model Prompting: A Survey.

[code]

New Trends for Modern Machine Translation with Large Reasoning Models.

Are Your LLMs Capable of Stable Reasoning?

[code]

From Medprompt to o1: Exploration of Run-Time Strategies for Medical Challenge Problems and Beyond.

To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoning.

Can LLMs Generate Novel Research Ideas? A Large-Scale Human Study with 100+ NLP Researchers.

A Peek into Token Bias: Large Language Models Are Not Yet Genuine Reasoners.

[code]

Iteration Head: A Mechanistic Study of Chain-of-Thought

Do Large Language Models Latently Perform Multi-Hop Reasoning?

Premise Order Matters in Reasoning with Large Language Models.

The Impact of Reasoning Step Length on Large Language Models.

Large Language Models Cannot Self-Correct Reasoning Yet.

At Which Training Stage Does Code Data Help LLM Reasoning?

Measuring Faithfulness in Chain-of-Thought Reasoning.

Faith and Fate: Limits of Transformers on Compositionality.

Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting.

[code]

A Multitask, Multilingual, Multimodal Evaluation of ChatGPT on Reasoning, Hallucination, and Interactivity.

Large Language Models Can Be Easily Distracted by Irrelevant Context.

by Google Research et al., 2023 - Adding the instruction "Feel free to ignore irrelevant information given in the questions." consistently improves robustness to irrelevant context.

On Second Thought, Let's Not Think Step by Step! Bias and Toxicity in Zero-Shot Reasoning.

Bias and Toxicity in Zero-Shot Reasoning.

Towards Understanding Chain-of-Thought Prompting: An Empirical Study of What Matters.

[code]

Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Them.

[code]

Emergent Abilities of Large Language Models.

[blog]

Can language models learn from explanations in context?

JudgeLRM: Large Reasoning Models as a Judge.

Dynamic Benchmarking of Reasoning Capabilities in Code Large Language Models Under Data Contamination.

[code]

CRANE: Reasoning with constrained LLM generation.

Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketching.

[code]

Self-rewarding correction for mathematical reasoning.

Competitive Programming with Large Reasoning Models.

s1: Simple test-time scaling.

(2025) - small open reasoning recipe via budget-forcing.

DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning.

[project]

Towards System 2 Reasoning in LLMs: Learning How to Think With Meta Chain-of-Thought.

HuatuoGPT-o1, Towards Medical Complex Reasoning with LLMs

[code]

PPM: Automated Generation of Diverse Programming Problems for Benchmarking Code Generation Models

[code] Simin Chen, XiaoNing Feng, Xiaohong Han, Cong Liu, Wei Yang FSE'24

DRT-o1: Optimized Deep Reasoning Translation via Long Chain-of-Thought.

[code]

MALT: Improving Reasoning with Multi-Agent LLM Training.

SmartAgent: Chain-of-User-Thought for Embodied Personalized Agent in Cyber World.

Marco-o1: Towards Open Reasoning Models for Open-Ended Solutions.

[code] [model]

Embedding Self-Correction as an Inherent Ability in Large Language Models for Enhanced Mathematical Reasoning.

Deliberate Reasoning for LLMs as Structure-aware Planning with Accurate World Model.

[code]

Interpretable Contrastive Monte Carlo Tree Search Reasoning.

Training Language Models to Self-Correct via Reinforcement Learning.

OpenAI o1.

(2024-2025) - test-time-compute reasoning systems.

Agent Q: Advanced Reasoning and Learning for Autonomous AI Agents.

DotaMath: Decomposition of Thought with Code Assistance and Self-correction for Mathematical Reasoning.

[code]

LLM-ARC: Enhancing LLMs with an Automated Reasoning Critic.

Q*: Improving Multi-step Reasoning for LLMs with Deliberative Planning.

Buffer of Thoughts: Thought-Augmented Reasoning with Large Language Models.

[code]

Toward Self-Improvement of LLMs via Imagination, Searching, and Criticizing.

Self-playing Adversarial Language Game Enhances LLM Reasoning.

Evaluating Mathematical Reasoning Beyond Accuracy.

Advancing LLM Reasoning Generalists with Preference Trees.

LLM3: Large Language Model-based Task and Motion Planning with Motion Failure Reasoning.

[code]

Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking.

GLoRe: When, Where, and How to Improve LLM Reasoning via Global and Local Refinements.

Chain-of-Thought Reasoning Without Prompting.

V-STaR: Training Verifiers for Self-Taught Reasoners.

InternLM-Math: Open Math Large Language Models Toward Verifiable Reasoning.

Self-Discover: Large Language Models Self-Compose Reasoning Structures.

DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models.

K-Level Reasoning with Large Language Models.

Efficient Tool Use with Chain-of-Abstraction Reasoning.

Teaching Language Models to Self-Improve through Interactive Demonstrations.

Enhancing Zero-Shot Chain-of-Thought Reasoning in Large Language Models through Logic.

[code]

Chain-of-Verification Reduces Hallucination in Large Language Models.

Skeleton-of-Thought: Large Language Models Can Do Parallel Decoding.

Question Decomposition Improves the Faithfulness of Model-Generated Reasoning.

[code]

Let's Verify Step by Step.

process-supervised reward models for reasoning.

REFINER: Reasoning Feedback on Intermediate Representations.

[project] [code]

Active Prompting with Chain-of-Thought for Large Language Models.

[code]

Language Models as Inductive Reasoners.

Boosting LLM Reasoning: Push the Limits of Few-shot Learning with Reinforced In-Context Pruning.

Logic-LM: Empowering Large Language Models with Symbolic Solvers for Faithful Logical Reasoning.

[code]

Recursion of Thought: A Divide and Conquer Approach to Multi-Context Reasoning with Language Models.

[code] [poster]

Reasoning with Language Model is Planning with World Model.

Reasoning Implicit Sentiment with Chain-of-Thought Prompting.

[code]

Tree of Thoughts: Deliberate Problem Solving with Large Language Models.

[code]

SatLM: Satisfiability-Aided Language Models Using Declarative Prompting.

[code]

ART: Automatic multi-step reasoning and tool-use for large language models.

Automatic Prompt Augmentation and Selection with Chain-of-Thought from Labeled Data.

[code]

Synthetic Prompting: Generating Chain-of-Thought Demonstrations for Large Language Models.

Generating Chain-of-Thought Demonstrations for Large Language Models.

Faithful Chain-of-Thought Reasoning.

Rethinking with Retrieval: Faithful Large Language Model Inference.

by University of Pennsylvania et al., 2022 - They shows the potential of enhancing LLMs by retrieving relevant external knowledge based on decomposed reasoning steps obtained through chain-of-thought (CoT) prompting. I predict we're going to see many of these types of retrieval-enhanced LLMs in…

LAMBADA: Backward Chaining for Automated Reasoning in Natural Language.

Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions.

[code]

Large Language Models are Reasoners with Self-Verification.

[code]

Can Retriever-Augmented Language Models Reason? The Blame Game Between the Retriever and the Language Model.

[code]

Complementary Explanations for Effective In-Context Learning.

Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks.

[code]

Unsupervised Explanation Generation via Correct Instantiations.

PAL: Program-aided Language Models.

[project] [code]

Solving Math Word Problems via Cooperative Reasoning induced Language Models.

[code]

Large Language Models Can Self-Improve.

Mind's Eye: Grounded language model reasoning through simulation.

Automatic Chain of Thought Prompting in Large Language Models.

[code]

Language Models are Multilingual Chain-of-Thought Reasoners.

Ask Me Anything: A simple strategy for prompting language models.

[code]

Dynamic Prompt Learning via Policy Gradient for Semi-structured Mathematical Reasoning.

[project] [code]

Making Large Language Models Better Reasoners with Step-Aware Verifier.

Least-to-most prompting enables complex reasoning in large language models.

Self-consistency improves chain of thought reasoning in language models.

Multi-path sampling + majority vote: GSM8K 57% → 74%

Retrieval Augmentation for Commonsense Reasoning: A Unified Approach.

[code]

Language Models of Code are Few-Shot Commonsense Learners.

[code]

Solving Quantitative Reasoning Problems with Language Models.

[blog]

Large Language Models Still Can't Plan.

[code]

Large Language Models are Zero-Shot Reasoners.

"Let's think step by step" — zero-shot CoT milestone

Iteratively Prompt Pre-trained Language Models for Chain of Thought.

[code]

Chain of Thought Prompting Elicits Reasoning in Large Language Models.

[blog]

Introducing Visual Perception Token into Multimodal Large Language Model.

[code] [model] [dataset]

LlamaV-o1: Rethinking Step-by-step Visual Reasoning in LLMs.

[project] [code] [model]

Embodied-Reasoner: Synergizing Visual Search, Reasoning, and Action for Embodied Interactive Tasks

[project][code][dataset] Wenqi Zhang, Mengna Wang, Gangao Liu, Xu Huixin, Yiwei Jiang, Yongliang Shen, Guiyang Hou, Zhe Zheng, Hang Zhang, Xin Li, Weiming Lu, Peng Li, Yueting Zhuang. Preprint'25

Insight-V: Exploring Long-Chain Visual Reasoning with Multimodal Large Language Models.

[code] [model]

LLaVA-CoT: Let Vision Language Models Reason Step-by-Step

code model

Visual Sketchpad: Sketching as a Visual Chain of Thought for Multimodal Language Models.

[project] [code]

Chart-based Reasoning: Transferring Capabilities from LLMs to VLMs.

SpatialVLM: Endowing Vision-Language Models with Spatial Reasoning Capabilities.

[project]

Chain-of-Table: Evolving Tables in the Reasoning Chain for Table Understanding.

Link-Context Learning for Multimodal LLMs.

[code]

Gemini in Reasoning: Unveiling Commonsense in Multimodal Large Language Models.

G-LLaVA: Solving Geometric Problems with Multi-Modal Large Language Model.

Chameleon: Plug-and-Play Compositional Reasoning with Large Language Models.

[project] [code]

MM-REACT: Prompting ChatGPT for Multimodal Reasoning and Action.

[project] [code] [demo]

ViperGPT: Visual Inference via Python Execution for Reasoning.

[project] [code]

Visual ChatGPT: Talking, Drawing and Editing with Visual Foundation Models.

[code]

Multimodal Chain-of-Thought Reasoning in Language Models.

[code]

Visual Programming: Compositional Visual Reasoning without Training.

[project] [code]

Socratic Models: Composing Zero-Shot Multimodal Reasoning with Language.

[project] [code]

Learning to Reason from Feedback at Test-Time.

[code]

S²R: Teaching LLMs to Self-verify and Self-correct via Reinforcement Learning.

[code]

rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinking.

[code]

MathScale: Scaling Instruction Tuning for Mathematical Reasoning.

Learning Deductive Reasoning from Synthetic Corpus based on Formal Logic.

[code]

Symbolic Chain-of-Thought Distillation: Small Models Can Also "Think" Step-by-Step.

[code]

Specializing Smaller Language Models towards Multi-Step Reasoning.

Large Language Models Are Reasoning Teachers.

[code]

Teaching Small Language Models to Reason.

They finetune a student model on the chain of thought (CoT) outputs generated by a larger teacher model. For example, the accuracy of T5 XXL on GSM8K improves from 8.11% to 21.99% when finetuned on PaLM-540B generated chains of thought.

Distilling Multi-Step Reasoning Capabilities of Large Language Models into Smaller Models via Semantic Decompositions.

Scaling Instruction-Finetuned Language Models.

by Google - They find that instruction finetuning with the above aspects dramatically improves performance on a variety of model classes (PaLM, T5, U-PaLM), prompting setups (zero-shot, few-shot, CoT), and evaluation benchmarks. Flan-PaLM 540B achieves SoTA performance on several benchmarks. They…

Agent Shadow Brain

Self-evolving AI coding intelligence with infinite memory (TurboQuant), genetic algorithm self-evolution, predictive bug detection, PageRank knowledge graphs, swarm intelligence, and adversarial defense.

LLM Reasoners

A library for advanced large language model reasoning.

Chain-of-Thought Hub

Benchmarking LLM reasoning performance with chain-of-thought prompting.

Omni Skills Forge

50,000+ curated AI agent skills for Claude Code, Cursor, Copilot, Windsurf, Cline. Visual dashboard, one-click install, skill doctor, auto-update.

ThoughtSource

Central and open resource for data and tools related to chain-of-thought reasoning in large language models.

AgentChain

Chain together LLMs for reasoning & orchestrate multiple large models for accomplishing complex tasks.

google/Cascades

Python library which enables complex compositions of language models such as scratchpads, chain of thought, tool use, selection-inference, and more.

LogiTorch

PyTorch-based library for logical reasoning on natural language.

salesforce/LAVIS

One-stop Library for Language-Vision Intelligence.

facebookresearch/RAM

A framework to study AI models in Reasoning, Alignment, and use of Memory (RAM).