👍 83
11/10 08:00
Weight-space programming, introduced in recent MIT work by Takehi et al., enables a frontier LLM agent to read a network's weights and write a repair. Their GPT-6 Astra agent corrected the reasoning loops of Qwen3.5-0.8B without gradients, using 158 agent turns, 89.7 million tokens, and 50 forward-p
中文介绍 提出 Tree-Ising 方法,通过比特级 Ising 编辑修复语言模型。在无需梯度的情况下,利用 GPT-6 Astra 智能体读取网络权重并执行修复,成功纠正 Qwen3.5-0.8B 的推理循环,耗时 158 个智能体轮次、8970 万 token 和 50 次前向传播。
👍 83
11/10 08:00
Few-step sampling reduces the cost of streaming video generation. Distribution Matching Distillation (DMD) supervises few-step generators with a separate fake-score network, which adds training memory and computation and must track the evolving generator. We extend APEX to streaming video by using c
中文介绍 将 APEX 扩展到流式视频生成,通过条件移位实现少步采样,避免 DMD 中额外 fake-score 网络带来的内存与计算开销。在保持生成质量的同时显著减少采样步数,降低流式生成成本。
👍 66
11/09 08:00
Trained action policies write manipulation skill into network weights, tying what is learned to the distribution it is learned from. In-context learning (ICL) avoids this coupling but has not matched trained policies, because a demonstration recorded in another scene carries no feedback from the wor
中文介绍 提出世界上下文学习(World ICL)方法,让机器人操作策略通过演示上下文获取反馈,避免将技能固化于网络权重。该方法使智能体在未见场景中能根据世界反馈调整行为,缩小了与训练策略的性能差距。
👍 61
11/10 08:00
External memory enables LLM agents to improve from experience by distilling trajectories of past tasks into concise lessons for new tasks. However, memory can also lead to memory-induced degradation, where an agent fails on tasks it would solve without memory, often because it follows a lesson that
中文介绍 针对 LLM 智能体外部记忆导致性能退化的问题,提出证据链接记忆(Evidence-Linked Memory)方法,将记忆与具体证据关联。该方法能有效避免智能体盲目遵循错误经验,减少记忆诱导退化,提升新任务成功率。
👍 35
11/08 08:00
Generation, imputation, and prediction query different aspects of the same tabular distribution, but are usually evaluated separately. This leaves unclear which capabilities they share, how their training objectives interact, and whether one model can perform all three well. We introduce the One Tab
中文介绍 构建 One Table 基准,首次统一评估表格模型在生成、插补和预测三种任务上的能力。发现三种任务共享底层分布信息,但训练目标存在冲突,为开发通用表格模型提供实证依据。
👍 33
11/09 08:00
Discrete autoregressive modeling, notably through next-token prediction, has become a cornerstone of modern language models. Extending this paradigm to vision allows visual and language models to converge on a unified framework, thereby fostering shared methodology and infrastructure. This work inve
中文介绍 提出 dNaViT 框架,将离散自回归建模(如 next-token prediction)扩展到视觉领域,使视觉和语言模型统一于同一范式。实验表明离散编码可作为通用视觉表示,提升下游任务性能并促进跨模态共享基础设施。
👍 26
10/29 08:00
Mathematical supervision for language models is typically flattened into isolated problem-solution pairs, discarding the curricular organization and evidential structure of the original sources. We introduce TOME-Math (Textbook-Oriented Mathematical Evidence), a large-scale corpus that converts 2,11
中文介绍 提出 TOME-Math 语料库,从 211 本结构化教科书中提取数学证明监督信号,保留课程组织与证据结构。该语料库可用于训练语言模型的数学推理能力,优于传统扁平化问题-答案对。
👍 26
11/09 08:00
Early successes on agentic robotics point to a potential paradigm shift toward using coding agents to control robots. Existing robotics benchmarks are built around visuomotor policy learning, leaving it open whether coding agents can reliably control diverse robots under physical-world constraints.
中文介绍 提出 Embodied First Exam 基准,首次系统评估编码智能体在物理世界约束下控制多种机器人的能力。发现现有编码智能体在实时性、鲁棒性和物理一致性方面存在明显不足,为具身智能研究提供新测试标准。
👍 26
11/10 08:00
System-1 Decision Models directly make structured probabilistic decisions over query-specified options given a state in a single forward pass, without autoregressively generating reasoning trajectories or textual responses. Recent evaluations cover diverse tasks but often rely on existing benchmarks
中文介绍 提出 MMDM 基准,用于系统评估系统-1 决策模型(一次性前向输出结构化概率决策,无自回归推理)的能力。覆盖多种任务,揭示现有评估偏倚,为开发高效决策模型提供更全面的测试框架。
👍 23
11/05 08:00
Language agents increasingly work for users on the open Web, and finance is where trusting them is hardest. The data must come from the Web, the strategy lives in the user's head, and correctness depends on time, so an agent can fail silently: a strategy the user never asked for, unverified data, or
中文介绍 提出 FINCOSMOS 开放世界评估范式,用于金融领域语言智能体。该范式强调数据来自真实网络、策略源于用户、正确性依赖时间,能捕捉智能体在金融场景中的静默失败(如策略未请求、数据未验证),提供更可靠的评测框架。
👍 22
11/07 08:00
Weight-activation quantization can reduce the memory and compute cost of neural-network inference, but recent methods often require expensive optimization and architecture-specific propagation of basis transforms across layers. We introduce MACaQ, an MXFP4 W4A4 post-training quantization method that
中文介绍 提出 MACaQ,一种模型无关的校准后训练量化方法,用于低精度权重与激活(W4A4 MXFP4)。该方法避免昂贵的优化和跨层基变换传播,能快速实现低精度推理,显著降低内存与计算开销。
👍 17
11/07 08:00
How little of a pretrained large language model (LLM) has to change for it to acquire new capabilities? We find that zeroing fewer than 0.05% of its weights, with no other modification, matches full fine-tuning (FFT) at scale across three LLM post-training regimes: reinforcement learning with verifi
中文介绍 发现零化预训练 LLM 中少于 0.05% 的权重即可匹配全量微调效果,覆盖 RLVR、指令微调和偏好优化三种后训练场景。该结果表明模型能力获取可依赖极少数权重的定向修改,极大降低后训练成本。
👍 15
10/29 08:00
Large language models (LLMs) draw on parametric knowledge and retrieval-augmented generation (RAG) to solve knowledge-intensive tasks. Yet problem solving in unfamiliar domains calls for tightly coupled knowledge acquisition and reasoning, beyond factual retrieval and aggregation or learning solely
中文介绍 提出通过强化学习激励 LLM 的自我学习能力,使模型在未知领域能自主获取知识并结合推理解决知识密集型任务。该方法超越单纯 RAG 或参数化知识,实现知识获取与推理的紧密耦合,提升泛化性能。
👍 11
11/09 08:00
AAA games offer a rich testbed for evaluating agent perception, reasoning, and control. However, evaluating whether agents can complete tasks in AAA games is challenging and often human-annotated because task outcomes are often difficult to verify automatically, let alone applying it for Reinforceme
中文介绍 提出 GTAGym,首个基于真实 AAA 游戏的任务环境,用于评估和强化学习智能体。该环境提供自动任务验证机制,支持强化学习训练,克服了传统人工标注成本高的难题,为复杂游戏场景中的智能体研究提供平台。
👍 11
11/01 11:58
A scientific agent can give the same verdict while reading and citing different evidence. BioAgent-DF tests this separation using paired reference and candidate values saved at the same scientific computation site. It restores or transplants those values at their original coordinates, then runs the
中文介绍 提出 BioAgent-DF 测试框架,审计生物医学研究智能体的决策来源。通过配对参考值与候选值在原始坐标的恢复或移植,检测智能体是否引用了不相关证据却给出相同结论,揭示科学智能体的决策偏差。
👍 8
11/07 08:00
Diffusion language models (DLMs) generate text through iterative denoising, where decisions about which tokens to commit or remask strongly affect output quality. Existing certainty-aware strategies rely on instantaneous confidence or entropy, but these local signals can be misleading on difficult r
中文介绍 提出频率感知重掩码(Frequency-Aware Remasking)策略用于扩散语言模型测试时扩展。该方法考虑 token 重掩码的全局频率信息,克服传统置信度或熵的局部信号在困难推理任务中的误导性,提升生成质量。
👍 7
11/06 08:00
Large language models (LLMs) increasingly support Web-based services that must balance response quality and inference cost across heterogeneous requests. Conventional routing selects one model per prompt, while fixed ensembles cannot adapt their composition and cost to individual requests. We introd
中文介绍 提出集成路由方法,针对提示自适应且成本感知的 LLM 推理。根据每个请求的动态需求组合多个模型并优化推理成本,避免固定集成无法适应个体请求的问题,在保持质量的同时降低推理开销。
👍 7
11/04 08:00
We introduce Micius-Flash, a 27-billion-parameter model for engineering with native multimodal input, deep reasoning and a 256K-token context window. We design the Micius data pipeline as a knowledge-tree-driven data flywheel comprising three stages: knowledge tree construction and expansion, data s
中文介绍 发布 Micius-Flash,270 亿参数的工程领域多模态模型,支持原生多模态输入、深度推理和 256K 上下文。通过知识树驱动的数据飞轮(知识树构建、数据筛选、合成)实现高效训练,在工程任务上表现优异。
👍 7
11/10 08:00
Memory-augmented vision-language-action (VLA) policies are meant to remember what happened. Many of them, however, fill memory by time: they sample frames uniformly across the episode and stamp each frame with its time step. Because time passes even when nothing happens, such a memory may not tell w
中文介绍 发现记忆增强 VLA 策略存在时间捷径:许多方法按时间均匀采样并打上时间戳,导致记忆仅反映时间流逝而非事件发生。提出改进方法避免此类捷径,提升机器人记忆的真实性和有效性。
👍 5
11/10 08:00
Foundation models such as GPT and Claude have shown strong capabilities in spatial understanding and embodied intelligence. These advances make agentic robotics a promising path toward artificial general intelligence in robotics. However, we find that using foundation models to control robots poses
中文介绍 发现将基础模型(如 GPT、Claude)用于机器人控制存在安全隐患。提出 LEASH 框架,通过安全约束和监控机制确保智能体在物理世界中的行为安全,为具身智能的可靠部署提供保障。