This repository contains the resources for Test-time Computing: from System-1 Thinking to System-2 Thinking

• Test-time Adaptation
  • Updating the Model
  • Modifying the Input
  • Editing the Representation
  • Calibrating the Output
• Test-time Reasoning
  • Feedback Modeling
    • Score-based
    • Verbal-based
  • Search Strategies
    • Repeated sampling
    • Self-correction
    • Tree search
• Future Directions
  • Generalizable System-2 Model
  • Multimodal Reasoning
  • Efficiency and Performance Trade-off
  • Scaling Law
  • Strategy Combination

• Test-Time Training with Self-Supervision for Generalization under Distribution Shifts [ICML 2020] paper
• MT3: Meta Test-Time Training for Self-Supervised Test-Time Adaption [AISTATS 2022] paper code
• Test-Time Training with Masked Autoencoders [NeurIPS 2022] paper
• TTT++: When Does Self-Supervised Test-Time Training Fail or Thrive? [NeurIPS 2021] paper code
• Efficient Test-Time Prompt Tuning for Vision-Language Models [arxiv 2024.8] paper
• Tent: Fully Test-time Adaptation by Entropy Minimization [ICLR 2021] paper code
• MEMO: Test Time Robustness via Adaptation and Augmentation [NeurIPS 2022] paper code
• The Entropy Enigma: Success and Failure of Entropy Minimization [arxiv 2024.5] paper code
• On Pitfalls of Test-Time Adaptation [ICML 2023] paper code
• Beware of Model Collapse! Fast and Stable Test-time Adaptation for Robust Question Answering [EMNLP 2023] paper code
• Align Your Prompts: Test-Time Prompting with Distribution Alignment for Zero-Shot Generalization [NeurIPS 2024] paper
• Protected Test-Time Adaptation via Online Entropy Matching: A Betting Approach [arxiv 2024.8] paper code
• Simulating Bandit Learning from User Feedback for Extractive Question Answering [ACL 2022] paper code
• Using Interactive Feedback to Improve the Accuracy and Explainability of Question Answering Systems Post-Deployment [ACL 2022] paper
• Test-time Adaptation for Machine Translation Evaluation by Uncertainty Minimization [ACL 2023] paper code
• COMET: A Neural Framework for MT Evaluation [EMNLP 2020] paper
• Test-Time Adaptation with CLIP Reward for Zero-Shot Generalization in Vision-Language Models [ICLR 2023] paper code
• Improving robustness against common corruptions by covariate shift adaptation [NeurIPS 2020] paper
• Selective Annotation Makes Language Models Better Few-Shot Learners [arxiv 2022.9] paper code
• Test-Time Adaptation with Perturbation Consistency Learning [arxiv 2023.4] paper
• Test-Time Prompt Adaptation for Vision-Language Models [NeurIPS 2023] paper
• Diverse Data Augmentation with Diffusions for Effective Test-time Prompt Tuning [ICCV 2023] paper code
• Test-Time Model Adaptation with Only Forward Passes [ICML 2024] paper code
• Test-Time Low Rank Adaptation via Confidence Maximization for Zero-Shot Generalization of Vision-Language Models [arxiv 2024.7] paper code
• StreamAdapter: Efficient Test Time Adaptation from Contextual Streams [arxiv 2024.11] paper
• Towards Stable Test-time Adaptation in Dynamic Wild World [ICLR 2023] paper code
• SoTTA: Robust Test-Time Adaptation on Noisy Data Streams [NeurIPS 2023] paper code
• Robust Question Answering against Distribution Shifts with Test-Time Adaption: An Empirical Study [EMNLP 2022] paper code

• What Makes Good In-Context Examples for GPT-3? [DeeLIO 2022] paper
• In-Context Learning with Iterative Demonstration Selection [EMNLP 2024] paper
• Dr.ICL: Demonstration-Retrieved In-context Learning [arxiv 2023.5] paper
• Learning To Retrieve Prompts for In-Context Learning [NAACL 2022] paper
• Unified Demonstration Retriever for In-Context Learning [ACL 2023] paper code
• Why Can GPT Learn In-Context? Language Models Secretly Perform Gradient Descent as Meta-Optimizers [ACL 2023] paper code
• Finding Support Examples for In-Context Learning [EMNLP 2023] paper code
• Large Language Models Are Latent Variable Models: Explaining and Finding Good Demonstrations for In-Context Learning [NeurIPS 2023] paper code
• Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivity [ACL 2022] paper
• Self-Adaptive In-Context Learning: An Information Compression Perspective for In-Context Example Selection and Ordering [ACL 2023] paper
• RetICL: Sequential Retrieval of In-Context Examples with Reinforcement Learning [arxiv 2024.4] paper
• Automatic Chain of Thought Prompting in Large Language Models [ICLR 2022] paper code
• Self-ICL: Zero-Shot In-Context Learning with Self-Generated Demonstrations [EMNLP 2023] paper code
• Z-ICL: Zero-Shot In-Context Learning with Pseudo-Demonstrations [ACL 2023] paper
• Self-Generated In-Context Learning: Leveraging Auto-regressive Language Models as a Demonstration Generator [arxiv 2022.6] paper
• Demonstration Augmentation for Zero-shot In-context Learning [ACL 2024] paper code

• Plug and Play Language Models: A Simple Approach to Controlled Text Generation [ICLR 2022] paper
• Steering Language Models With Activation Engineering [arxiv 2024.10] paper
• Improving Instruction-Following in Language Models through Activation Steering [arxiv 2024.10] paper
• Inference-Time Intervention: Eliciting Truthful Answers from a Language Model [arxiv 2024.6] paper code
• Refusal in Language Models Is Mediated by a Single Direction [arxiv 2024.10] paper code
• In-context Vectors: Making In Context Learning More Effective and Controllable Through Latent Space Steering [arxiv 2024.2] paper code
• Investigating Bias Representations in Llama 2 Chat via Activation Steering [arxiv 2024.2] paper
• Personalized Steering of Large Language Models: Versatile Steering Vectors Through Bi-directional Preference Optimization [arxiv 2024.7] paper code
• Spectral Editing of Activations for Large Language Model Alignment [NeurIPS 2024] paper code
• Multi-property Steering of Large Language Models with Dynamic Activation Composition [BlackboxNLP 2024] paper code

• Generalization through Memorization: Nearest Neighbor Language Models [ICLR 2020] paper code
• Nearest Neighbor Machine Translation [ICLR 2021] paper code
• Efficient Cluster-Based k-Nearest-Neighbor Machine Translation [ACL 2022] paper code
• What Knowledge Is Needed? Towards Explainable Memory for kNN-MT Domain Adaptation [ACL 2023] paper code
• Efficient Domain Adaptation for Non-Autoregressive Machine Translation [ACL 2024] paper code
• kNN-NER: Named Entity Recognition with Nearest Neighbor Search [arxiv 2022.3] paper code
• kNN-CM: A Non-parametric Inference-Phase Adaptation of Parametric Text Classifiers [EMNLP 2023] paper code
• AdaNPC: Exploring Non-Parametric Classifier for Test-Time Adaptation [ICML 2023] paper code

• Scaling Laws for Reward Model Overoptimization [PMLR 2023] paper
• Training Verifiers to Solve Math Word Problems [arxiv 2021.10] paper
• Advancing LLM Reasoning Generalists with Preference Trees [arxiv 2024.4] paper code
• V-STaR: Training Verifiers for Self-Taught Reasoners [COLM 2024] paper
• Solving math word problems with process- and outcome-based feedback [arxiv 2022.11] paper
• Let's Verify Step by Step [ICLR 2024] paper code
• Entropy-Regularized Process Reward Model arxiv 2024.12 code
• Math-Shepherd: Verify and Reinforce LLMs Step-by-step without Human Annotations [ACL 2024] paper
• Improve Mathematical Reasoning in Language Models by Automated Process Supervision [arxiv 20224.6] paper
• Rewarding Progress: Scaling Automated Process Verifiers for LLM Reasoning [arxiv 20224.10] paper
• AutoPSV: Automated Process-Supervised Verifier [NeurIPS 2024] paper code
• Free Process Rewards without Process Labels [arxiv 2024.12] paper code
• Critique-out-Loud Reward Models [arxiv 2024.8] paper code
• Improving Reward Models with Synthetic Critiques [arxiv 2024.5] paper
• Generative Verifiers: Reward Modeling as Next-Token Prediction [arxiv 2024.8] paper
• Self-Generated Critiques Boost Reward Modeling for Language Models [arxiv 2024.11] paper

• Is ChatGPT a Good NLG Evaluator? A Preliminary Study [ACL 2023] paper code
• ChatGPT as a Factual Inconsistency Evaluator for Text Summarization [arxiv 2023.3] paper
• G-Eval: NLG Evaluation using Gpt-4 with Better Human Alignment [ACL 2023] pdf code
• Can Large Language Models Be an Alternative to Human Evaluations? [ACL 2023] paper
• LLMs instead of Human Judges? A Large Scale Empirical Study across 20 NLP Evaluation Tasks [arxiv 2024.6] paper
• Large Language Models are not Fair Evaluators [ACL 2024] paper code
• Large Language Models are Inconsistent and Biased Evaluators [arxiv 2024.5] paper
• Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena [NeurIPS 2023] paper code
• PandaLM: An Automatic Evaluation Benchmark for LLM Instruction Tuning Optimization [ICLR 2024] paper
• JudgeLM: Fine-tuned Large Language Models are Scalable Judges [arxiv 2023.10] paper code
• Fennec: Fine-grained Language Model Evaluation and Correction Extended through Branching and Bridging [arxiv 2024.5] paper code
• REFINER: Reasoning Feedback on Intermediate Representations [ACL 2024] paper
• Shepherd: A Critic for Language Model Generation [arxiv 2023.8] paper code
• Generative Judge for Evaluating Alignment [ICLR 2024] paper code
• Can LLMs Produce Faithful Explanations For Fact-checking? Towards Faithful Explainable Fact-Checking via Multi-Agent Debate [ICLR 2024] paper

• Competition-level code generation with alphacode [Science 2022] paper code
• Code Llama: Open Foundation Models for Code [arxiv 2023.8] paper code
• More Agents Is All You Need [arxiv 2024.2] paper code
• Just Ask One More Time! Self-Agreement Improves Reasoning of Language Models in (Almost) All Scenarios [ACL 2024] paper
• Self-Consistency Improves Chain of Thought Reasoning in Language Models [ICLR 2023] paper
• Not All Votes Count! Programs as Verifiers Improve Self-Consistency of Language Models for Math Reasoning[arxiv 2024.10] paper code
• Learning to summarize with human feedback[NeurIPS 2020] paper
• Training Verifiers to Solve Math Word Problems[arxiv 2021.10] paper
• WebGPT: Browser-assisted question-answering with human feedback [arxiv 2021.12] paper
• Making Language Models Better Reasoners with Step-Aware Verifier [ACL 2023] paper code
• Accelerating Best-of-N via Speculative Rejection [ICML 2024] paper
• TreeBoN: Enhancing Inference-Time Alignment with Speculative Tree-Search and Best-of-N Sampling [arxiv 2024.10] paper
• Fast Best-of-N Decoding via Speculative Rejection [NeurIPS 2024] paper
• Adaptive Inference-Time Compute: LLMs Can Predict if They Can Do Better, Even Mid-Generation [arxiv 2024.10] paper code
• Preference-Guided Reflective Sampling for Aligning Language Models [EMNLP 2024] paper code
• Reinforced Self-Training (ReST) for Language Modeling[arxiv 2023.8] paper
• Variational Best-of-N Alignment [arxiv 2024.7] paper
• BoNBoN Alignment for Large Language Models and the Sweetness of Best-of-n Sampling [NeurIps 2024] paper
• BOND: Aligning LLMs with Best-of-N Distillation [arxiv 2024.7] paper
• Inference-Aware Fine-Tuning for Best-of-N Sampling in Large Language Models [arxiv 2024.12] paper

• Reflexion: Language Agents with Verbal Reinforcement Learning [arxiv 2023.3] paper code
• Interscript: A dataset for interactive learning of scripts through error feedback [arxiv 2021.12] paper code
• NL-EDIT: Correcting Semantic Parse Errors through Natural Language Interaction [ACL 2021] paper code
• Learning to repair: Repairing model output errors after deployment using a dynamic memory of feedback [ACL 2022] paper code
• CRITIC: Large Language Models Can Self-Correct with Tool-Interactive Critiquing [ICLR 2024] paper code
• Teaching Large Language Models to Self-Debug [ICLR 2024] paper
• RARR: Researching and Revising What Language Models Say, Using Language Models [ACL 2023] paper code
• Graph-based, Self-Supervised Program Repair from Diagnostic Feedback [ICML 2020] paper
• Improving Factuality and Reasoning in Language Models through Multiagent Debate [arxiv 2023.5] paper code
• Examining Inter-Consistency of Large Language Models Collaboration: An In-depth Analysis via Debate [EMNLP 2023] paper code
• Encouraging Divergent Thinking in Large Language Models through Multi-Agent Debate [EMNLP 2024] paper code
• ReConcile: Round-Table Conference Improves Reasoning via Consensus among Diverse LLMs [ACL 2024] paper code
• Mixture-of-Agents Enhances Large Language Model Capabilities [arxiv 2024.6] paper code
• Can LLMs Produce Faithful Explanations For Fact-checking? Towards Faithful Explainable Fact-Checking via Multi-Agent Debate [arxiv 2024.7] paper
• Debating with More Persuasive LLMs Leads to More Truthful Answers [ICML 2024] paper code
• ChatEval: Towards Better LLM-based Evaluators through Multi-Agent Debate [ICLR 2024] pdf
• ChainLM: Empowering Large Language Models with Improved Chain-of-Thought Prompting [IREC 2024] paper
• Are You Sure? Challenging LLMs Leads to Performance Drops in The FlipFlop Experiment [arxiv 2023.11] paper
• MultiAgent Collaboration Attack: Investigating Adversarial Attacks in Large Language Model Collaborations via Debate [arxiv 2024.6] paper
• Teaching Models to Balance Resisting and Accepting Persuasion [arxiv 2024.10] paper code
• GroupDebate: Enhancing the Efficiency of Multi-Agent Debate Using Group Discussion [arxiv 2024.9] paper
• Improving Multi-Agent Debate with Sparse Communication Topology [arxiv 2024.6] paper
• Self-Rewarding Language Models [arxiv 2024.1] paper
• Constitutional AI: Harmlessness from AI Feedback [arxiv 2022.12] paper code
• Self-Refine: Iterative Refinement with Self-Feedback [NeurIPS 2023] paper
• Language Models can Solve Computer Tasks [arxiv 2023.3] paper code
• Confidence Matters: Revisiting Intrinsic Self-Correction Capabilities of Large Language Models [arxiv 2024.2] paper code
• Is Self-Repair a Silver Bullet for Code Generation? [ICLR 2024] paper code
• Large Language Models Cannot Self-Correct Reasoning Yet [ICLR 2024] paper
• Reasoning in Token Economies: Budget-Aware Evaluation of LLM Reasoning Strategies [arxiv 2024.6] paper
• Can Large Language Models Really Improve by Self-critiquing Their Own Plans? [arxiv 2023.10] paper
• GPT-4 Doesn't Know It's Wrong: An Analysis of Iterative Prompting for Reasoning Problems [arxiv 2023.10] paper
• When Can LLMs Actually Correct Their Own Mistakes? A Critical Survey of Self-Correction of LLMs [arxiv 2024.6] paper
• LLMs cannot find reasoning errors, but can correct them given the error location [ACL 2024] paper code
• Self-critiquing models for assisting human evaluators [arxiv 2022.6] paper
• Recursive Introspection: Teaching Language Model Agents How to Self-Improve [arxiv 2024.7] paper
• Embedding Self-Correction as an Inherent Ability in Large Language Models for Enhanced Mathematical Reasoning [arxiv 2024.10] paper
• Learn Beyond The Answer: Training Language Models with Reflection for Mathematical Reasoning [arxiv 2024.6] paper code
• GLoRe: When, Where, and How to Improve LLM Reasoning via Global and Local Refinements [arxiv 2024.2] paper
• Generating Sequences by Learning to Self-Correct [ICLR 2023] paper code
• Training Language Models to Self-Correct via Reinforcement Learning [arxiv 2024.9] paper

• Tree of Thoughts: Deliberate Problem Solving with Large Language Models [NeurIPS 2023] paper code
• Self-Evaluation Guided Beam Search for Reasoning [NeurIPS 2023] paper code
• Reasoning with Language Model is Planning with World Model [EMNLP 2023] paper code
• Accessing GPT-4 level Mathematical Olympiad Solutions via Monte Carlo Tree Self-refine with LLaMa-3 8B [arxiv 2024.6] paper code
• Forest-of-Thought: Scaling Test-Time Compute for Enhancing LLM Reasoning [arxiv 2025.2] paper code
• Large Language Model Guided Tree-of-Thought [arxiv 2023.5] paper code
• Reasoning with Language Model is Planning with World Model [EMNLP 2023] paper code
• Alphazero-like Tree-Search can Guide Large Language Model Decoding and Training [arxiv 2023.9] paper
• Mutual Reasoning Makes Smaller LLMs Stronger Problem-Solvers [arxiv 2024.8] paper code
• Interpretable Contrastive Monte Carlo Tree Search Reasoning [arxiv 2024.10] paper code
• ReST-MCTS: LLM Self-Training via Process Reward Guided Tree Search [arxiv 2024.6] pdf code
• Monte Carlo Tree Search Boosts Reasoning via Iterative Preference Learning [arxiv 2024.5] paper code
• O1 Replication Journey: A Strategic Progress Report -- Part 1 [arxiv 2024.10] paper code
• Marco-o1: Towards Open Reasoning Models for Open-Ended Solutions [arxiv 2024.11] paper code
• o1-Coder: an o1 Replication for Coding [arxiv 2024.12] paper code

• A Baseline Analysis of Reward Models’ Ability To Accurately Analyze Foundation Models Under Distribution Shift [arxiv 2023.10] paper
• Evaluating Robustness of Reward Models for Mathematical Reasoning [arxiv 2024.10] paper
• ODIN: Disentangled Reward Mitigates Hacking in RLHF [ICML 2024] paper
• Interpretable Preferences via Multi-Objective Reward Modeling and Mixture-of-Experts [arxiv 2024.6] paper code
• DogeRM: Equipping Reward Models with Domain Knowledge through Model Merging [EMNLP 2024] paper code
• Regularizing Hidden States Enables Learning Generalizable Reward Model for LLMs [NeurIPS 2024] paper code
• Generalizing Reward Modeling for Out-of-Distribution Preference Learning [ECML-PKDD 2024] paper code
• Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision [arxiv 2023.12] paper
• MAF: Multi-Aspect Feedback for Improving Reasoning in Large Language Models [EMNLP 2023] paper code
• Training language models to follow instructions with human feedback [arxiv 2022.3] paper

• Test-Time Adaptation with CLIP Reward for Zero-Shot Generalization in Vision-Language Models [ICLR 2024] paper code
• Multimodal Chain-of-Thought Reasoning in Language Models [TMLR 2024] paper code
• Mind's Eye of LLMs: Visualization-of-Thought Elicits Spatial Reasoning in Large Language Models [NeurIPS 2024] paper
• KAM-CoT: Knowledge Augmented Multimodal Chain-of-Thoughts Reasoning [AAAI 2024] paper
• Multimodal Reasoning with Multimodal Knowledge Graph [ACL 2024] paper
• Interleaved-Modal Chain-of-Thought [arxiv 2024.11] paper
• LLaVA-Critic: Learning to Evaluate Multimodal Models [arxiv 2024.10] paper
• LLaVA-CoT: Let Vision Language Models Reason Step-by-Step [arxiv 2024.11] paper code

• Learning How Hard to Think: Input-Adaptive Allocation of LM Computation [arxiv 2024.10] paper
• Scaling LLM Inference with Optimized Sample Compute Allocation [arxiv 2024.10] paper code
• Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs [arxiv 2025.2] paper
• Reasoning in Token Economies: Budget-Aware Evaluation of LLM Reasoning Strategies [EMNLP 2024] paper
• Prometheus: Inducing Fine-Grained Evaluation Capability in Language Models [ICLR 2024] paper
• MoDE-CoTD: Chain-of-Thought Distillation for Complex Reasoning Tasks with Mixture of Decoupled LoRA-Experts [ACL 2024] paper
• Quantization Meets Reasoning: Exploring LLM Low-Bit Quantization Degradation for Mathematical Reasoning [arxiv 2025.2] paper
• LazyLLM: Dynamic Token Pruning for Efficient Long Context LLM Inference [arxiv 2024.7] paper
• Token-Budget-Aware LLM Reasoning [arxiv 2024.12] paper code
• Fast Inference from Transformers via Speculative Decoding [ICML 2022] paper
• Compressed Chain of Thought: Efficient Reasoning Through Dense Representations [arxiv 2024.12] paper
• Unlocking Efficiency in Large Language Model Inference: A Comprehensive Survey of Speculative Decoding [arxiv 2024.6] paper code

• Large Language Monkeys: Scaling Inference Compute with Repeated Sampling [arxiv 2024.9] paper
• Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters [arxiv 2024.8] paper
• Inference Scaling Laws: An Empirical Analysis of Compute-Optimal Inference for Problem-Solving with Language Models [arxiv 2024.10] paper code
• A Simple and Provable Scaling Law for the Test-Time Compute of Large Language Models [arxiv 2024.11] paper

• The Surprising Effectiveness of Test-Time Training for Abstract Reasoning [arxiv 2024.11] paper code
• Marco-o1: Towards Open Reasoning Models for Open-Ended Solutions [arxiv 2024.11] paper code
• Beyond Examples: High-level Automated Reasoning Paradigm in In-Context Learning via MCTS [arxiv 2024.11] paper

If our survey is helpful to your research, please cite our paper:

@article{ji2025test,
  title={Test-time Computing: from System-1 Thinking to System-2 Thinking},
  author={Ji, Yixin and Li, Juntao and Ye, Hai and Wu, Kaixin and Xu, Jia and Mo, Linjian and Zhang, Min},
  journal={arXiv preprint arXiv:2501.02497},
  year={2025}
}