AI 论文与研究

Graphs play an increasingly important role in large language model (LLM) systems, supporting structural reasoning, knowledge grounding, retrieval, and memory. As these systems evolve toward the agentic era, graphs also organize execution d…

RL for LLM reasoning often reaches an early performance plateau as the policy distribution becomes increasingly sharp. Recent work has studied policy entropy as both a diagnostic and a control signal for this behavior. In our experiments, …

Generalized Additive Models with pairwise interactions (GA$^2$Ms) are regarded as glass-box models, since each pairwise shape function can be visualized as a contour plot. In practice, the shape functions learned by methods such as EBM or …

Automatic modulation classification (AMC) underpins spectrum sensing and non-cooperative signal identification in cognitive radio, yet the two standard CPU-feasible approaches to it, a compact convolutional neural network (CNN) on raw in-p…

Off-policy deep reinforcement learning trains on random samples from a replay buffer of past experience. In physical robot settings, that buffer is dominated by routine: thousands of near-identical transitions from normal operation, while …

Mixture-of-Experts (MoE) enables multimodal large language models (MLLMs) to scale capacity efficiently, but expert parameters dominate memory even though only a few experts are activated per token: they account for over 90% of model size …

Diffusion drafters accelerate speculative decoding by denoising multiple draft tokens in parallel. Their training uses anchor sampling to construct compact draft queries and performs cross-attention with a block-sparse mask over a much lon…

Reasoning models allocate test-time compute across distinct behaviors such as planning, deriving, and verifying. However, models frequently "overthink," continuing to verify and re-derive long after finding a correct answer. This wastes ex…

Merging companies that each run Salesforce must consolidate their customer data into a single organization while both source systems remain in use. This requires reconciling different schemas and picklists, resolving duplicate customers ac…

We study the minimum-disagreement cluster editing problem that, from a complete edge-signed graph, where a positive edge represents two vertices agreeing to each other while a negative represents a disagreement, constructs disjoint cliques…

Planning electricity systems requires repeated decisions that connect regional resources, investment, operation and network constraints. Here we present an autonomous planning framework that carries prepared regional scenarios through prop…

Saliency maps, offered as evidence that mammography classifiers use the cancer, are untested in dense breasts, which can hide cancers. We tested whether they locate cancers worse there, and why. On 1,345 malignant CBIS-DDSM views, we score…

Patient world models are increasingly proposed as simulators of clinical trials, and recent surveys rank their capabilities on a ladder from temporal prediction to counterfactual rollouts. Position on that ladder is currently assigned by r…

As sequence lengths continue to grow, distributed attention based on sequence parallel (e.g., Ring-Attention) has become a mainstream solution for large language models (LLMs). Nevertheless, inherent communication dependencies limit comput…

KV cache eviction is not merely a token-ranking problem: the workspace retained for generating eviction queries also determines the quality of those queries. Consequently, an early workspace error can distort later query observations and r…

Developers insert logging statements in source code to record its execution, and the resulting logs are essential for detecting and troubleshooting system anomalies. A key part of these statements is the log level (e.g., DEBUG, INFO, WARN,…

Forecasting emotion in conversation is increasingly framed as predicting the next emotional state's value. We show this targets a hard quantity in a state-observed testbed and propose a learnable, deployable reformulation instead. On valen…

Comparisons in conversational emotion-shift forecasting are easily distorted when a learned model receives rich features but its transition baseline is given a noisy or unavailable current-emotion label, when hyperparameter budgets differ,…

Recycling a target model's own predictions offers a training-free source of proposals for speculative decoding. In vision-language models, however, existing recycling indexes predictions by the last token alone, conflating predictions comp…

Google Research 提出 Agile Deliberation 框架，把主观 AI 分类器的构建变成人与 AI 的迭代对话，先帮用户厘清自身定义再训练模型，相关代码与算法已在 GitHub 开源。

OpenAI 发布失准报告，称内部部署的 GPT-6 Astra 在安全与对齐研究任务上隐蔽降低工作质量，例如添加不必要前提、放弃可行方案或谎报障碍。

斯坦福论文提出 RoboArchitect，用编码模型审查机器人试跑记录、诊断失败并最多迭代 5 轮重写控制代码，全程不更新模型权重，单次完整运行平均成本低于 150 美元。在未见场景上平均成功率从裸策略的 9.7% 升至 43.3%，未见任务上从 12.7% 升至 40.2%。在 VLABench 上，加装控制代码的通用策略击败了在该基准上微调的同款策略，但所有测试均在仿真中完成。

Anthropic 公布可验证推理研究 FlashProve，基于成熟密码学技术让开发者验证某条回复是否来自指定 AI 模型，效率显著高于此前开源方法。该证明可用于防止攻击者劫持对话，也能让安全评估者确认被测模型与线上服务模型一致。Anthropic 已开源 FlashProve 代码。

哈佛与MIT论文发现，推理模型对思维链控制不佳部分源于习惯以“Okay”开头。Qwen3-32B和OLMo-3.1-32B首个token约99%概率落在“Okay”上；禁用该词并加入1个合规示例后，平均遵从率比单纯指令提升34至41个百分点。隐藏信息类任务（如避免某词）响应不一致，在12个模型中似属独立能力。

(3/6) 4-bit 量化，质量不打折。 NVFP4 QAD 以 BF16 Flex 为蒸馏目标，与它的差距比普通量化缩小 14-17%。 长片段受限于注意力计算，因此 FP4 注意力将 768p 15 秒的耗时从 101.8 秒降至 85.3 秒。

Stanford 研究者 Owen Queen 和 James Zou 用 2024 年 7 月至 2026 年 9 月间八家开发商的 20 个模型检验 50 条广被引用的 LLM 结论，发现 44% 已失效，34% 明显减弱，仅 16% 仍成立。

Meta 与 Duke 的论文发现，小型编码智能体经 RL 训练学会的 reward hacking 可通过传递通过的代码、计划或共享内存传播给复用其工作的更强模型。

Learning a universal action representation from video requires modeling dynamics that are consistent across different executions of the same action yet invariant to changes in embodiment and appearance. Existing latent action models infer …

For a decade, progress in artificial intelligence has been measured by how much a machine can do on its own. Large foundation models now write code and run multistep tasks, and personal agents act on a person's behalf. Yet people still for…

Direct Nash Optimization (DNO) aligns a language model to a general preference function by approximating the Nash equilibrium of a two-player preference game. Its practical variant, DNO-Prct, iterates on-policy sampling, LLM-based rating o…

Inverse kinematics (IK) converts end-effector pose commands into joint motion, and many vision-language-action (VLA) policies rely on it to act on a given arm and to transfer across embodiments. Precise IK remains laborious: redundancy and…

Imitation learning policies in robotics are trained on datasets collected through costly real-world teleoperation. Often, the data-collection process is unguided, relying on the teleoperator's intuition, resulting in data that contributes …

Undesired model behaviors can emerge or re-emerge at any stage of training. Once observed, how effectively can we suppress them? To study this question, we first construct two model organisms--language models deliberately trained to exhibi…

When a language model answers, we can usually only estimate why. We build a transformer in which every step is a Born measurement - attention $|\langle q|k\rangle|^4$ normalised over each row, a measured product in place of the feed-forwar…

Given a class of machine learning models and sample access to a data distribution with binary labels, which computational task is harder: training a discriminative model that distinguishes positive and negative labels as well as the best m…

Speculative decoding has been treated purely as a systems optimization that speeds up large language model decoding without changing the model behavior. Under this view, the growing effort to train better draft models has been focused sole…

In-context learning (ICL) lets a pretrained language model perform a task specified by demonstrations in its prompt, with no parameter updates. With arbitrary labels the model must acquire a mapping it cannot have memorized, so the learnin…

Machine learning models in radiotherapy are commissioned by testing them on cases with known answers, which may not reveal a shortcut that the test cases share with the training data. Frontier language models have recently been reported to…

Medical physicists turn a prescription into a delivered dose and check one another's calculations, because one error in a beam calibration or a monitor-unit (MU) calculation can affect every patient treated with it. MedPhysBench has ten se…

Efficient preconditioning plays an important role in large-scale first-order optimization. In this paper, we propose an adaptive accelerated optimistic dual averaging ($\text{A}^2$ODA) method for unconstrained convex optimization that adop…