mlsanityData to 14 Oct 2026

The Surprising Effectiveness of Test-Time Training for Few-Shot Learning

arXiv 2411.07279 ↗First posted 11 Nov 2024cs.AI cs.CL cs.LGcited by 96

Not on the map. Why?

Cited by 30 in LLM reasoning, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

Language models (LMs) have shown impressive performance on tasks within their training distribution, but often struggle with structurally novel tasks even when given a small number of in-context task examples. We investigate the effectiveness of test-time training (TTT) -- temporarily updating model parameters during inference using a loss derived from input data -- as a mechanism for improving LMs' reasoning and few-shot learning capabilities. On the Abstraction and Reasoning Corpus (ARC), performing TTT with in-context examples yields up to 6× higher accuracy compared to fine-tuned baselines -- reaching 53.0% on the public validation set with an 8B-parameter LM and 61.9% when ensembled with program-synthesis methods, matching average human performance. On BIG-Bench Hard (BBH), TTT on in-context examples surpasses standard few-shot prompting in the 10-shot setting by 7.3 percentage points (50.5% to 57.8%). Our findings highlight the limitations of in-context learning for novel tasks and demonstrate the potential of test-time training to enhance language model adaptability.

Key facts. Cited by 96 papers in the map (data through 14 Oct 2026). Builds on 31 of its 43 references that are in the map. First posted on arXiv 11 Nov 2024.

Builds on

31 of 43 references.

References · 31

  1. 2020Language Models are Few-Shot Learnerscited by 32,419
  2. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  3. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  4. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,708
  5. 2024GPT-4o System Cardcited by 4,588
  6. 2023QLoRA: Efficient Finetuning of Quantized LLMscited by 3,418
  7. 2023Tree of Thoughts: Deliberate Problem Solving with Large Language Modelscited by 3,211
  8. 2022Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language modelscited by 1,998
  9. 2017Decoupled Weight Decay Regularizationcited by 1,831
  10. 2022Challenging BIG-Bench Tasks and Whether Chain-of-Thought Can Solve Themcited by 1,807
  11. 2022Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?cited by 1,434
  12. 2024Large Language Monkeys: Scaling Inference Compute with Repeated Samplingcited by 991
19 more
  1. 2022What learning algorithm is in-context learning? Investigations with linear modelscited by 684
  2. 2021MetaICL: Learning to Learn In Contextcited by 454
  3. 2024Learning to (Learn at Test Time): RNNs with Expressive Hidden Statescited by 355
  4. 2023Function Vectors in Large Language Modelscited by 308
  5. 2023Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Taskscited by 246
  6. 2022Test-Time Training with Masked Autoencoderscited by 212
  7. 2023Hypothesis Search: Inductive Reasoning with Language Modelscited by 120
  8. 2023Test-Time Training on Nearest Neighbors for Large Language Modelscited by 88
  9. 2024Learning How Hard to Think: Input-Adaptive Allocation of LM Computationcited by 70
  10. 2024Combining Induction and Transduction for Abstract Reasoningcited by 53
  11. 2024Efficiently Learning at Test-Time: Active Fine-Tuning of LLMscited by 42
  12. 2021Communicating Natural Programs to Humans and Machinescited by 35
  13. 2024Addressing the Abstraction and Reasoning Corpus via Procedural Example Generationcited by 33
  14. 2024H-ARC: A Robust Estimate of Human Performance on the Abstraction and Reasoning Corpus Benchmarkcited by 33
  15. 2024CodeIt: Self-Improving Language Models with Prioritized Hindsight Replaycited by 23
  16. 2024Probing the Decision Boundaries of In-context Learning in Large Language Modelscited by 17
  17. 2024Neural networks for abstraction and reasoning: Towards broad generalization in machinescited by 13
  18. 2023Solving ARC visual analogies with neural embeddings and vector arithmetic: A generalized methodcited by 8
  19. 2024Do Large Language Models Solve ARC Visual Analogies Like People Do?cited by 7

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 40; 2026: 56.

96 papers.

Sort:

By area:

  1. 2026Reinforcement Learning via Self-Distillationcited by 332
  2. 2025Self-Adapting Language Modelscited by 84
  3. 2025Nested Learning: The Illusion of Deep Learning Architecturescited by 72
  4. 2025Aristotle: IMO-level Automated Theorem Provingcited by 58
  5. 2025Training a Scientific Reasoning Model for Chemistrycited by 56
  6. 2025Jigsaw-R1: A Study of Rule-based Visual Reinforcement Learning with Jigsaw Puzzlescited by 36
  7. 2025Self-Improving LLM Agents at Test-Timecited by 30
  8. 2025ArcMemo: Abstract Reasoning Composition with Lifelong LLM Memorycited by 24
  9. 2026SIA: Self Improving AI with Harness & Weight Updatescited by 21
  10. 2025Test-Time Adaptation for LLM Agents via Environment Interactioncited by 20
  11. 2026AdaJEPA: An Adaptive Latent World Modelcited by 20
  12. 2025FineMedLM-o1: Enhancing Medical Knowledge Reasoning Ability of LLM from Supervised Fine-Tuning to Test-Time Trainingcited by 19
38 more
  1. 2026TTCS: Test-Time Curriculum Synthesis for Self-Evolvingcited by 19
  2. 2025SLOT: Sample-specific Language Model Optimization at Test-timecited by 18
  3. 2025Let's (not) just put things in Context: Test-Time Training for Long-Context LLMscited by 18
  4. 2026Back into Plato's Cave: Examining Cross-modal Representational Convergence at Scalecited by 16
  5. 2025Making Small Language Models Efficient Reasoners: Intervention, Supervision, Reinforcementcited by 14
  6. 2025Reconstruct, Inpaint, Test-Time Finetune: Dynamic Novel-view Synthesis from Monocular Videoscited by 14
  7. 2025DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learningcited by 12
  8. 2025Certified Self-Consistency: Statistical Guarantees and Test-Time Training for Reliable Reasoning in LLMscited by 12
  9. 2026Spatial-TTT: Streaming Visual-based Spatial Intelligence with Test-Time Trainingcited by 11
  10. 2025Learning to Reason from Feedback at Test-Timecited by 10
  11. 2025ARC-GEN: A Mimetic Procedural Benchmark Generator for the Abstraction and Reasoning Corpuscited by 10
  12. 2025Local Mixtures of Experts: Essentially Free Test-Time Training via Model Mergingcited by 9
  13. 2025Learning with Calibration: Exploring Test-Time Computing of Spatio-Temporal Forecastingcited by 9
  14. 2025LatentEvolve: Self-Evolving Test-Time Scaling in Latent Spacecited by 9
  15. 2025Learning on the Job: Test-Time Curricula for Targeted Reinforcement Learningcited by 9
  16. 2025Shoot First, Ask Questions Later? Building Rational Agents that Explore and Act Like Peoplecited by 9
  17. 2026Tool Verification for Test-Time Reinforcement Learningcited by 9
  18. 2025From Generation to Generalization: Emergent Few-Shot Learning in Video Diffusion Modelscited by 8
  19. 2025Test-time Offline Reinforcement Learning on Goal-related Experiencecited by 8
  20. 2025MiGrATe: Mixed-Policy GRPO for Adaptation at Test-Timecited by 7
  21. 2025Diverse Inference and Verification for Advanced Reasoningcited by 6
  22. 2025Test-Time Matching: Unlocking Compositional Reasoning in Multimodal Modelscited by 6
  23. 2025You only need 4 extra tokens: Synergistic Test-time Adaptation for LLMscited by 6
  24. 2025Specialization after Generalization: Towards Understanding Test-Time Training in Foundation Modelscited by 5
  25. 2026Consensus as Privileged Context for Label-Free Self-Distillationcited by 5
  26. 2025Consistency-based Abductive Reasoning over Perceptual Errors of Multiple Pre-trained Models in Novel Environmentscited by 4
  27. 2025RTTC: Reward-Guided Collaborative Test-Time Computecited by 4
  28. 2025CausalARC: Abstract Reasoning with Causal World Modelscited by 4
  29. 2026Hair-Trigger Alignment: Black-Box Evaluation Cannot Guarantee Post-Update Alignmentcited by 4
  30. 2026TTT-Parkour: Rapid Test-Time Training for Perceptive Robot Parkourcited by 4
  31. 2026No Time Like the Present: Agentic Test-Time Training for LLM Agentscited by 4
  32. 2026AgentStream: How Well Do Self-Evolving LLM Agents Perform Under Streaming Tasks?cited by 4
  33. 2025Context Tuning for In-Context Optimizationcited by 3
  34. 2025Test time training enhances in-context learning of nonlinear functionscited by 3
  35. 2025ARC-AGI Without Pretrainingcited by 3
  36. 2026ReasonCACHE: Teaching LLMs To Reason Without Weight Updatescited by 3
  37. 2026Meta-TTL: Meta-Learning Self-Improvement Policies for Language Agentscited by 3
  38. 2026One Step Forward and K Steps Back: Better Reasoning with Denoising Recursion Modelscited by 3

50 most cited of 96 All 96 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026FewMMBench: A Benchmark for Multimodal Few-Shot Learning
  2. 2025Less is More: Improving LLM Reasoning with Minimal Test-Time Intervention
  3. 2024Many-Shot In-Context Learning
  4. 2026Not the Example, but the Process: How Self-Generated Examples Enhance LLM Reasoning
  5. 2026Beyond Perplexity: A Behavioral Evaluation Framework for Deployment-Memory Claims in LLM Test-Time Training
  6. 2022Can In-context Learners Learn a Reasoning Concept from Demonstrations?
  7. 2026CL-bench: A Benchmark for Context Learning
  8. 2024Leveraging Zero-Shot Prompting for Efficient Language Model Distillation