mlsanityData to 7 Oct 2026

From Few to Many: Self-Improving Many-Shot Reasoners Through Iterative Optimization and Generation

arXiv 2502.00330 ↗First posted 1 Feb 2025cs.LG cs.AI stat.MLcited by 20

On the map: LLM reasoning → Reasoning benchmarks and evaluation

also in 1 more
Abstract (arXiv)

Recent advances in long-context large language models (LLMs) have led to the emerging paradigm of many-shot in-context learning (ICL), where it is observed that scaling many more demonstrating examples beyond the conventional few-shot setup in the context can lead to performance benefits. However, despite its promise, it is unclear what aspects dominate the benefits and whether simply scaling to more examples is the most effective way of improving many-shot ICL. In this work, we first provide an analysis of the factors driving many-shot ICL, and we find that 1) many-shot performance can still be attributed to often a few disproportionately influential examples and 2) identifying such influential examples ("optimize") and using them as demonstrations to regenerate new examples ("generate") can lead to further improvements. Inspired by the findings, we propose BRIDGE, an algorithm that alternates between the optimize step with Bayesian optimization to discover the influential sets of examples and the generate step to reuse this set to expand the reasoning paths of the examples back to the many-shot regime automatically. On Gemini, Claude, and Mistral LLMs of different sizes, we show that BRIDGE to significant improvements across a diverse set of tasks, including symbolic reasoning, numerical reasoning, and code generation.

Key facts. Cited by 20 papers in the map (data through 7 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 50 of its 72 references that are in the map. First posted on arXiv 1 Feb 2025.

Builds on

50 of 72 references.

Closest ancestors · 1

Same direction, newest first

  1. 2022PAL: Program-aided Language Models18 Nov

Other references · 49

  1. 2020Language Models are Few-Shot Learnerscited by 32,199
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,084
  3. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,690
  4. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,203
  5. 2017Axiomatic Attribution for Deep Networkscited by 3,350
  6. 2024Gemini 1.5: Unlocking multimodal understanding across millions of tokens of contextcited by 3,175
  7. 2023Mistral 7Bcited by 3,065
  8. 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 2,674
  9. 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,348
  10. 2021Calibrate Before Use: Improving Few-Shot Performance of Language Modelscited by 1,371
  11. 2021What Makes Good In-Context Examples for GPT-3?cited by 1,184
  12. 2021Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivitycited by 1,122
37 more
  1. 2022STaR: Bootstrapping Reasoning With Reasoningcited by 895
  2. 2023DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelinescited by 803
  3. 2018A Tutorial on Bayesian Optimizationcited by 734
  4. 2022Automatic Chain of Thought Prompting in Large Language Modelscited by 684
  5. 2023Can LLM Already Serve as A Database Interface? A BIg Bench for Large-Scale Database Grounded Text-to-SQLscited by 667
  6. 2021Learning To Retrieve Prompts for In-Context Learningcited by 641
  7. 2024Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Surveycited by 641
  8. 2018GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Accelerationcited by 603
  9. 2024LESS: Selecting Influential Data for Targeted Instruction Tuningcited by 514
  10. 2024Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programscited by 270
  11. 2016Professor Forcing: A New Algorithm for Training Recurrent Networkscited by 214
  12. 2024Many-Shot In-Context Learningcited by 203
  13. 2022Diverse Demonstrations Improve In-context Compositional Generalizationcited by 136
  14. 2024CHASE-SQL: Multi-Path Reasoning and Preference Optimized Candidate Selection in Text-to-SQLcited by 132
  15. 2024In-Context Learning with Long-Context Models: An In-Depth Explorationcited by 126
  16. 2024Retrieval Augmented Generation or Long-Context LLMs? A Comprehensive Study and Hybrid Approachcited by 117
  17. 2021Case-based Reasoning for Natural Language Queries over Knowledge Basescited by 107
  18. 2018A Flexible Framework for Multi-Objective Bayesian Optimization using Random Scalarizationscited by 98
  19. 2023Amortizing intractable inference in large language modelscited by 94
  20. 2023Batch Calibration: Rethinking Calibration for In-Context Learning and Prompt Engineeringcited by 91
  21. 2022Structured Prompting: Scaling In-Context Learning to 1,000 Examplescited by 81
  22. 2024Gecko: Versatile Text Embeddings Distilled from Large Language Modelscited by 81
  23. 2024In-context Learning with Retrieved Demonstrations for Language Models: A Surveycited by 69
  24. 2024Many-Shot In-Context Learning in Multimodal Foundation Modelscited by 64
  25. 2023Better Zero-Shot Reasoning with Self-Adaptive Promptingcited by 52
  26. 2021Think Global and Act Local: Bayesian Optimisation over High-Dimensional Categorical and Mixed Search Spacescited by 51
  27. 2022Bidirectional Language Models Are Also Few-shot Learnerscited by 48
  28. 2023Self-ICL: Zero-Shot In-Context Learning with Self-Generated Demonstrationscited by 42
  29. 2024Dual Operating Modes of In-Context Learningcited by 42
  30. 2023In-Context Learning with Many Demonstration Examplescited by 39
  31. 2022Bayesian Optimization over Discrete and Mixed Spaces via Probabilistic Reparameterizationcited by 35
  32. 2024Teach Better or Show Smarter? On Instructions and Exemplars in Automatic Prompt Optimizationcited by 28
  33. 2024Fairer Preferences Elicit Improved Human-Aligned Large Language Model Judgmentscited by 25
  34. 2021Unsupervised Neural Machine Translation with Generative Language Models Onlycited by 24
  35. 2025Reliable Text-to-SQL with Adaptive Abstentioncited by 19
  36. 2019Scalarizing Functions in Bayesian Multiobjective Optimizationcited by 15
  37. 2023Universal Self-Adaptive Promptingcited by 14

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 13; 2026: 7.

20 papers.

Sort:

By area:

  1. 2025AccelOpt: A Self-Improving LLM Agentic System for AI Accelerator Kernel Optimizationcited by 21
  2. 2025A Survey of Scaling in Large Language Model Reasoningcited by 10
  3. 2025RoT: Enhancing Table Reasoning with Iterative Row-Wise Traversalscited by 10
  4. 2025Linear-Time Demonstration Selection for In-Context Learning via Gradient Estimationcited by 8
  5. 2025Self-Improving Model Steeringcited by 3
  6. 2025Towards Compute-Optimal Many-Shot In-Context Learningcited by 3
  7. 2026Optimizing In-Context Demonstrations for LLM-based Automated Gradingcited by 3
  8. 2026Many-Shot CoT-ICL: Making In-Context Learning Truly Learncited by 3
  9. 2025Distilling Many-Shot In-Context Learning into a Cheat Sheetcited by 2
  10. 2025Agentic Policy Optimization via Instruction-Policy Co-Evolutioncited by 2
  11. 2025Towards more Contextual Agents: An extractor-Generator Optimization Frameworkcited by 1
  12. 2025Large Language Models are Demonstration Pre-Selectors for Themselvescited by 1
8 more
  1. 2025Learning-to-Context Slope: Evaluating In-Context Learning Effectiveness Beyond Performance Illusionscited by 1
  2. 2026AdapShot: Adaptive Many-Shot In-Context Learning with Semantic-Aware KV Cache Reusecited by 1
  3. 2025From Implicit Exploration to Structured Reasoning: Leveraging Guideline and Refinement for LLMscited by 0
  4. 2025COM-BOM: Bayesian Exemplar Search for Efficiently Exploring the Accuracy-Calibration Pareto Frontiercited by 0
  5. 2026Task-Aware Structured Memory for Dynamic Multi-modal In-Context Learningcited by 0
  6. 2026MILES: Modular Instruction Memory with Learnable Selection for Self-Improving LLM Reasoningcited by 0
  7. 2026Long-Context Demonstration Selection Using State Space Modelscited by 0
  8. 2026Efficient Task Adaptation in Large Language Models: A Survey of Weight-Based, Prompt-Based, and Embedding-Based Adaptationscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025MachineLearningLM: Scaling Many-shot In-context Learning via Continued Pretraining
  2. 2026Not the Example, but the Process: How Self-Generated Examples Enhance LLM Reasoning
  3. 2025MIR-Bench: Can Your LLM Recognize Complicated Patterns via Many-Shot In-Context Reasoning?
  4. 2025Refract ICL: Rethinking Example Selection in the Era of Million-Token Models
  5. 2023In-Context Learning with Iterative Demonstration Selection
  6. 2025Compressing Many-Shots in In-Context Learning
  7. 2024Does In-Context Learning Really Learn? Rethinking How Large Language Models Respond and Solve Tasks via In-Context Learning
  8. 2025What is an "Abstract Reasoner"? Revisiting Experiments and Arguments about Large Language Models