1
Relative Value Learning
ICLR 2026论文提出相对价值学习新范式，或为强化学习领域带来突破性进展
arXiv:2607.21120v1 Announce Type: cross Abstract: In reinforcement learning, critics typically estimate absolute state values $V(s)$, estimating how g…
2
PM-KVQ: Progressive Mixed-precision KV Cache Quantization for Long-CoT LLMs
针对长思维链大语言模型的KV缓存量化新突破，渐进混合精度方案，已被ICLR 2026接收，代码已开源。
arXiv:2505.18610v2 Announce Type: replace Abstract: Recently, significant progress has been made in developing reasoning-capable Large Language Models…
3
SGD-Based Knowledge Distillation with Bayesian Teachers: Theory and Guidelines
用贝叶斯教师指导SGD蒸馏，理论框架与实用指南双管齐下，读懂知识蒸馏新范式。
arXiv:2601.01484v2 Announce Type: replace Abstract: Knowledge Distillation (KD) is a central paradigm for transferring knowledge from a large teacher …
4
Spinning Straw into Gold: Relabeling LLM Agent Trajectories in Hindsight for Successful Demonstrations
让失败轨迹逆向重生，用后见之明把废料炼成黄金，有效提升LLM Agent成功率。
arXiv:2607.04235v1 Announce Type: new Abstract: Large language model agents operate in partially observable, long-horizon settings where obtaining sup…
5
Sakana Fugu: How Collaborative AI is Changing the Game
多模型协作框架Fugu，协调GPT-5、Claude等专家模型，比单一巨模型更高效灵活
# Sakana Fugu: The Multi-Agent AI System That Works Like a Team We’ve all been there: copy-pasting a prompt from ChatGPT to Claude, and then to Gemini…
6
Natural Identifiers for Privacy and Data Audits in Large Language Models
用自然标识符给大模型做隐私审计，全新视角兼顾数据保护与可追溯性，AI安全领域值得关注的前沿研究。
arXiv:2606.24408v1 Announce Type: new Abstract: Assessing the privacy of large language models (LLMs) presents significant challenges. In particular, …
7
Learning Patient-Specific Disease Dynamics with Latent Flow Matching for Longitudinal Imaging Generation
ICLR 2026最新研究，用潜流匹配模型生成纵向医学影像，精准捕捉患者病情动态变化。
arXiv:2512.09185v4 Announce Type: replace Abstract: Understanding disease progression is a central clinical challenge with direct implications for ear…
8
The Price of Agreement: Measuring LLM Sycophancy in Agentic Financial Applications
发现大模型在金融代理场景中为讨好用户而扭曲建议，一项测量迎合行为的新研究揭示潜在风险。
arXiv:2604.24668v3 Announce Type: replace Abstract: Given the increased use of LLMs in financial systems today, it becomes important to evaluate the s…
9
Difference-Aware Retrieval Policies for Imitation Learning
一键直达ICLR 2026最新模仿学习论文，包含代码和演示，助你快速掌握差异感知检索策略
arXiv:2606.09758v1 Announce Type: cross Abstract: Parametric imitation learning via behavior cloning can suffer from poor generalization to out-of-dis…
10
Enhancing Hallucination Detection through Noise Injection
揭秘ICLR 2026新方法：通过注入噪声提升大模型幻觉检测能力，思路新颖且有效。
arXiv:2502.03799v4 Announce Type: replace Abstract: Large Language Models (LLMs) are prone to generating plausible yet incorrect responses, known as h…
11
SoLoPO: Unlocking Long-Context Capabilities in LLMs via Short-to-Long Preference Optimization
SoLoPO提出短到长偏好优化，高效解锁大模型长上下文能力，被ICLR 2026录用。
arXiv:2505.11166v3 Announce Type: replace-cross Abstract: Despite advances in pretraining with extended context sizes, large language models (LLMs) st…
12
Quality-Diversity Evolution for Discovering Diverse Vulnerabilities in LLM Safety
用质量多样性进化算法系统性挖掘LLM安全漏洞，突破传统对抗测试的局限
arXiv:2606.00801v1 Announce Type: cross Abstract: Current approaches to LLM adversarial testing suffer from coverage gaps: manual red-teaming does not…
13
Uncertainty Estimation via Hyperspherical Confidence Mapping
ICLR 2026最新论文，提出超球面置信映射方法，高效提升深度学习模型的不确定性估计精度
arXiv:2605.05964v2 Announce Type: replace Abstract: Quantifying uncertainty in neural network predictions is essential for high-stakes domains such as…
14
Enhancing Visual Token Representations for Video Large Language Models via Training-Free Spatial-Temporal Pooling and Gridding
无需额外训练，通过空间-时间池化与网格化巧妙提升视频大语言模型视觉token表征，ICLR 2026接收！
arXiv:2605.22078v1 Announce Type: cross Abstract: Recent advances in Multimodal Large Language Models (MLLMs) have significantly advanced video unders…
15
Dr.LLM: Dynamic Layer Routing in LLMs
提出动态层路由机制，让LLM推理时跳过无关层，显著提升效率与精度。
arXiv:2510.12773v2 Announce Type: replace-cross Abstract: Large Language Models (LLMs) process every token through all layers of a transformer stack, …
16
Efficient Spatially-Variant Convolution via Differentiable Sparse Kernel Complex
提出DISK可微稀疏核复合体，实现高效空间可变卷积，已被ICLR 2026接收。
arXiv:2512.04556v2 Announce Type: replace-cross Abstract: Image convolution with complex kernels is a fundamental operation in photography, scientific…
17
Mixture-of-Experts Can Surpass Dense LLMs Under Strictly Equal Resource
MoE架构在严格等资源条件下首次证明超越稠密大模型，ICLR 2026最新研究。
arXiv:2506.12119v2 Announce Type: replace Abstract: Mixture-of-Experts (MoE) language models dramatically expand model capacity and achieve remarkable…
18
Eliminating Inductive Bias in Reward Models with Information-Theoretic Guidance
ICLR 2026 顶会论文：用信息论指导消除奖励模型中的归纳偏置，为强化学习对齐提供更客观的评估基础
arXiv:2512.23461v2 Announce Type: replace Abstract: Reward models (RMs) are essential in reinforcement learning from human feedback (RLHF) to align la…
19
Hybrid Training for Vision-Language-Action Models
ICLR 2026论文提出混合训练框架，统一视觉-语言-动作模型，提升多模态具身智能表现。
arXiv:2510.00600v2 Announce Type: replace-cross Abstract: Using Large Language Models to produce intermediate thoughts, a.k.a. Chain-of-thought (CoT),…
20
Proximal Diffusion Neural Sampler
ICLR 2026接受的论文，用近端优化改进扩散模型，实现更高效的神经采样器，适合机器学习研究者。
arXiv:2510.03824v2 Announce Type: replace Abstract: The task of learning a diffusion-based neural sampler for drawing samples from an unnormalized tar…