Can Test-Time Memory Evolution Outperform Test-Time Training for Scientific Discovery?
Summary
论文提出 Test-Time Memory Evolution（TTME），一种通过持续更新外部记忆而非模型权重来进行测试时自适应的方法，在五个数学任务上优于测试时训练（TTT），同时降低 82.8-84.6% 的 FLOPs 和 47.7-58.4% 的运行时间。
View Cached Full Text
Cached at: 10/28/26, 05:27 AM
Paper page - Can Test-Time Memory Evolution Outperform Test-Time Training for Scientific Discovery?
Source: https://huggingface.co/papers/2610.32046
Abstract
Test-Time Memory Evolution updates external memory without changing model weights to guide scientific discovery more efficiently than test-time training.
Recenttest-time training(TTT) frameworks have enabled open-weight large language models (LLMs) to outperform inference-only frameworks powered by frontier models on scientific discovery tasks. However, TTT introduces additional computational cost for online weight updates during the discovery process, making computational efficiency an important consideration when scaling to larger models. This motivates us to explore an efficienttest-time adaptationmethod for scientific discovery. We introduceTest-Time Memory Evolution(TTME), a plug-and-playtest-time adaptationmethod, as an alternative to TTT. TTME continually updates theexternal memoryby distilling the experience to guide subsequent discovery without updatingmodel weights. To have a clean comparison, we evaluate TTME and TTT under the same discovery framework, replacing only thetest-time adaptationmethod. Our evaluation shows that TTME outperforms TTT on all five mathematical tasks, while reducing computational cost by 82.8-84.6% inFLOPsand 47.7-58.4% in runtime. These results establish TTME as an effective and computationally efficient alternative to TTT in the evaluated discovery settings.
View arXiv pageView PDFGitHub1Add to collection
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2610.32046 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2610.32046 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2610.32046 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
EvoTest: Evolutionary Test-Time Learning for Self-Improving Agentic Systems
EvoTest introduces J-TTL, a benchmark for measuring agent test-time learning capabilities, and proposes an evolutionary framework where an Actor Agent plays games while an Evolver Agent iteratively improves the system's prompts, memory, and hyperparameters without fine-tuning. The method demonstrates superior performance compared to reflection and memory-based baselines on complex text-based games.
Self-Guided Test-Time Training for Long-Context LLMs
Proposes Self-Guided Test-Time Training (S-TTT), where the model identifies relevant evidence spans in long contexts for adaptation, achieving up to 15% relative improvement on long-context reasoning benchmarks.
TTPO: Test-Time Policy Optimization
TTPO introduces a label-free method for test-time training in mathematical reasoning that matches supervised performance by asymmetrically handling agreeing and disagreeing rollouts.
Scaling Self-Evolving Agents via Parametric Memory
Researchers from Alibaba/Qwen and Peking University introduce TMEM, a self-evolving parametric memory framework that uses online LoRA weight updates to let LLM agents genuinely learn from experience within a single episode, rather than relying solely on prompt-space memory. TMEM outperforms summary-based and retrieval-based baselines across multiple benchmarks including LoCoMo, LongMemEval-S, and CL-Bench.
Test-Time Memory Construction
提出 T2Mem 框架，将记忆构建延迟到推理阶段，随推理过程渐进式构建面向查询的记忆，在 LoCoMo 和 LongMemEval 上超越强基线并显著降低记忆构建开销。