Gamma-World: Generative Multi-Agent World Modeling Beyond Two Players
NVIDIA
Gamma-World：超越双参与者的生成式多智能体世界建模
World models for interactive video generation have largely focused on single-agent settings, where future observations are generated from a single control signal. However, many generated environments require multi-agent interaction: multiple players, robots, or embodied agents act simultaneously within a shared space. Scaling world models to such settings requires a principled multi-agent design: agents should remain independently controllable, permutation-symmetric, and support efficient inference while maintaining consistency across time and perspectives. In this paper, we present our generative multi-agent world model for interactive simulation. It introduces Simplex Rotary Agent Encoding, a parameter-free extension of 3D RoPE that represents agents as vertices of a regular simplex in rotary angle space. This gives each agent a distinct phase while making all agents permutation-equivalent, enabling scalable agent identity without learned per-slot identities or a fixed agent ordering. To avoid dense all-to-all attention across agents, we further propose Sparse Hub Attention, where learnable hub tokens mediate token interaction across agents, reducing cross-agent attention cost from quadratic to linear in the number of agents. For real-time rollout, we distill a full-context diffusion teacher into a causal student that generates temporal blocks sequentially with KV caching, enabling action-responsive generation at 24 FPS. Experiments in multiplayer virtual environments show that our model improves video fidelity, action controllability, and inter-agent consistency over slot-based and dense-attention baselines, while generalizing from two to four players without additional training.
用于交互式视频生成的世界模型主要聚焦于单智能体设置，其中未来观测由单个控制信号生成。然而，许多生成式环境需要多智能体交互：多名玩家、机器人或具身智能体在共享空间中同时行动。将世界模型扩展到此类设置需要具有明确原则的多智能体设计：智能体应保持可独立控制和置换对称，在维持跨时间、跨视角一致性的同时支持高效推理。本文提出用于交互式仿真的生成式多智能体世界模型。该模型引入单纯形旋转智能体编码，这是 3D RoPE 的一种无参数扩展，将智能体表示为旋转角空间中正单纯形的顶点。它为每个智能体赋予不同相位，同时使所有智能体在置换下等价，无需学习每个槽位的身份或固定智能体顺序即可扩展智能体身份表示。为避免智能体之间稠密的全对全注意力，我们进一步提出稀疏枢纽注意力，由可学习的枢纽 token 介导智能体间的 token 交互，将跨智能体注意力成本随智能体数量的增长从二次降至线性。为实现实时展开，我们将全上下文扩散教师模型蒸馏为因果学生模型，后者借助 KV cache 依次生成时间块，从而以 24 FPS 实现响应动作的生成。多玩家虚拟环境实验表明，与基于槽位和稠密注意力的基线相比，我们的模型提升了视频保真度、动作可控性和智能体间一致性，并且无需额外训练即可从两名玩家泛化到四名玩家。
MolmoAct2: Action Reasoning Models for Real-world Deployment
Ai2
MolmoAct2：面向真实世界部署的动作推理模型
Vision-Language-Action (VLA) models aim to provide a single generalist controller for robots, but today's systems fall short on the criteria that matter for real-world deployment. Frontier models are closed, open-weight alternatives are tied to expensive hardware, reasoning-augmented policies pay prohibitive latency for their grounding, and fine-tuned success rates remain below the threshold for dependable use. We present MolmoAct2, a fully open action reasoning model built for practical deployment, advancing its predecessor along five axes. We introduce MolmoER, a VLM backbone specialized for spatial and embodied reasoning, trained on a 3.3M-sample corpus with a specialize-then-rehearse recipe. We release three new datasets spanning low-to-medium cost platforms, including MolmoAct2-BimanualYAM, 720 hours of teleoperated bimanual trajectories that constitute the largest open bimanual dataset to date, together with quality-filtered Franka (DROID) and SO100/101 subsets. We provide OpenFAST, an open-weight, open-data action tokenizer trained on millions of trajectories across five embodiments. We redesign the architecture to graft a flow-matching continuous-action expert onto a discrete-token VLM via per-layer KV-cache conditioning. Finally, we propose MolmoThink, an adaptive-depth reasoning variant that re-predicts depth tokens only for scene regions that change between timesteps, retaining geometric grounding at a fraction of prior latency. In the most extensive empirical study of any open VLA to date, spanning 7 simulation and real-world benchmarks, MolmoAct2 outperforms strong baselines including Pi-05, while MolmoER surpasses GPT-5 and Gemini Robotics ER-1.5 across 13 embodied-reasoning benchmarks. We release model weights, training code, and complete training data. Project page: https://allenai.org/blog/molmoact2
视觉-语言-动作（VLA）模型旨在为机器人提供单一的通用控制器，但现有系统尚未达到真实世界部署所看重的标准。前沿模型并不开源；开放权重的替代方案依赖昂贵硬件；由推理增强的策略为实现定位付出了难以承受的延迟；经过微调的成功率也仍低于可靠使用所需的门槛。我们提出 MolmoAct2，一个为实际部署而构建的完全开放动作推理模型，从五个方面推进了其前身。我们引入 MolmoER，一个专门用于空间与具身推理的视觉语言模型骨干，采用“先专门化、后复习”的训练方案，在包含 330 万样本的语料库上训练。我们发布三个覆盖中低成本平台的新数据集，其中包括 MolmoAct2-BimanualYAM：它包含 720 小时遥操作双臂轨迹，是迄今规模最大的开放双臂数据集；此外还包括经过质量筛选的 Franka（DROID）和 SO100/101 子集。我们提供 OpenFAST，这是一个开放权重、开放数据的动作 tokenizer，在五种具身形态的数百万条轨迹上训练。我们重新设计架构，通过逐层 KV cache 条件化，将流匹配连续动作专家接入离散 token 视觉语言模型。最后，我们提出 MolmoThink，一种自适应深度推理变体，仅对相邻时间步之间发生变化的场景区域重新预测深度 token，从而以此前一小部分的延迟保留几何定位能力。在迄今针对开放 VLA 最全面的实证研究中，评测覆盖 7 个仿真与真实世界基准；MolmoAct2 超越包括 Pi-05 在内的强基线，MolmoER 则在 13 个具身推理基准上超过 GPT-5 和 Gemini Robotics ER-1.5。我们发布了模型权重、训练代码和完整训练数据。项目页面：https://allenai.org/blog/molmoact2
HRM-Text: Efficient Pretraining Beyond Scaling
Sapient AI
HRM-Text：超越规模扩展的高效预训练
The current pretraining paradigm for large language models relies on massive compute and internet-scale raw text, creating a significant barrier to foundational research. In contrast, biological systems demonstrate highly sample-efficient learning through multi-timescale processing, such as the functional organization of the frontoparietal loop. Taking this as inspiration, we introduce HRM-Text, which replaces standard Transformers with a Hierarchical Recurrent Model (HRM) that decouples computation into slow-evolving strategic and fast-evolving execution layers. To stabilize this deep recurrence for language modeling, we introduce MagicNorm and warmup deep credit assignment. Furthermore, instead of standard raw-text pretraining, we train exclusively on instruction-response pairs using a task-completion objective and PrefixLM masking. Serving as an empirical existence proof of efficient pretraining, a 1B-parameter HRM-Text model trained from scratch on only 40 billion unique tokens and $1,500 budget achieves 60.7% on MMLU, 81.9% on ARC-C, 82.2% on DROP, 84.5% on GSM8K, and 56.2% on MATH. Despite utilizing roughly 100-900x fewer training tokens and 96-432x less estimated compute than standard baselines, HRM-Text performs competitively with 2-7B parameter open models. These results demonstrate that co-designing architectures and objectives can radically reduce the compute-to-performance ratio, making pretraining from scratch accessible to the broader research community.
当前的大语言模型预训练范式依赖海量算力和互联网规模的原始文本，给基础研究设置了很高门槛。相比之下，生物系统能够通过多时间尺度处理实现极高的样本学习效率，例如额顶叶环路的功能组织。受此启发，我们提出 HRM-Text，以层级循环模型（HRM）取代标准 Transformer，将计算解耦为缓慢演化的策略层和快速演化的执行层。为稳定语言建模中的这种深层循环，我们引入 MagicNorm 和预热式深层信用分配。此外，我们不采用标准的原始文本预训练，而是仅在指令—回答对上使用任务完成目标和 PrefixLM 掩码进行训练。作为高效预训练确实可行的实证证明，一个从零开始训练的 10 亿参数 HRM-Text 模型仅使用 400 亿个不重复 token 和 1,500 美元预算，便在 MMLU、ARC-C、DROP、GSM8K 和 MATH 上分别达到 60.7%、81.9%、82.2%、84.5% 和 56.2%。尽管其训练 token 数量约比标准基线少 100 至 900 倍，估算算力少 96 至 432 倍，HRM-Text 的表现仍可与 20 亿至 70 亿参数的开放模型竞争。这些结果表明，协同设计架构与目标可以显著降低取得相同性能所需的计算量，使更广泛的研究群体能够开展从零开始的预训练。
CiteVQA: Benchmarking Evidence Attribution for Trustworthy Document Intelligence
OpenDataLab
CiteVQA：面向可信文档智能的证据归因基准
Multimodal Large Language Models (MLLMs) have significantly advanced document understanding, yet current Doc-VQA evaluations score only the final answer and leave the supporting evidence unchecked. This answer-only approach masks a critical failure mode: a model can land on the correct answer while grounding it in the wrong passage -- a critical risk in high-stakes domains like law, finance, and medicine, where every conclusion must be traceable to a specific source region. To address this, we introduce CiteVQA, a benchmark that requires models to return element-level bounding-box citations alongside each answer, evaluating both jointly. CiteVQA comprises 1,897 questions across 711 PDFs spanning seven domains and two languages, averaging 40.6 pages per document. To ensure fidelity and scalability, the ground-truth citations are generated by an automated pipeline-which identifies crucial evidence via masking ablation-and are subsequently validated through expert review. At the core of our evaluation is Strict Attributed Accuracy (SAA), which credits a prediction only when the answer and the cited region are both correct. Auditing 20 MLLMs reveals a pervasive Attribution Hallucination: models frequently produce the right answer while citing the wrong region. The strongest system (Gemini-3.1-Pro-Preview) achieves an SAA of only 76.0, and the strongest open-source MLLM reaches just 22.5. Ultimately, towards trustworthy document intelligence, CiteVQA exposes a reliability gap that answer-only evaluations overlook, providing the instrumentation needed to close it. Our repository is available at https://github.com/opendatalab/CiteVQA.
多模态大语言模型（MLLM）显著推进了文档理解，但当前的文档视觉问答评测只为最终答案评分，并不检查支撑答案的证据。这种仅评答案的方法掩盖了一种关键失败模式：模型可能得到正确答案，却将答案建立在错误段落之上。在法律、金融和医疗等高风险领域，这尤其危险，因为每个结论都必须能够追溯到具体来源区域。为此，我们提出 CiteVQA，一个要求模型在给出每个答案的同时返回元素级边界框引用，并对二者进行联合评测的基准。CiteVQA 包含来自 711 份 PDF 的 1,897 个问题，覆盖七个领域和两种语言，每份文档平均 40.6 页。为兼顾准确性与可扩展性，真实引用由自动化流水线生成；该流水线通过掩码消融识别关键证据，随后再由专家评审验证。评测的核心是严格归因准确率（SAA），只有答案和引用区域均正确时，预测才会得分。对 20 个多模态大语言模型的审计揭示了普遍存在的归因幻觉：模型经常回答正确，却引用错误区域。最强系统 Gemini-3.1-Pro-Preview 的 SAA 也只有 76.0，最强开源多模态大语言模型则仅为 22.5。最终，面向可信文档智能，CiteVQA 揭示了仅评答案的方法所忽视的可靠性差距，并提供了弥合这一差距所需的评测工具。我们的仓库位于 https://github.com/opendatalab/CiteVQA。
SkillOpt: Executive Strategy for Self-Evolving Agent Skills
Microsoft Research
SkillOpt：面向自演化智能体技能的执行策略
Agent skills today are hand-crafted, generated one-shot, or evolved through loosely controlled self-revision, none of which behaves like a deep-learning optimizer for the skill, and none of which reliably improves over its starting point under feedback. We argue the skill should instead be trained as the external state of a frozen agent, with the same discipline that makes weight-space optimization reproducible. SkillOpt is, to our knowledge, the first systematic controllable text-space optimizer for agent skills: a separate optimizer model turns scored rollouts into bounded add/delete/replace edits on a single skill document, and an edit is accepted only when it strictly improves a held-out validation score. A textual learning-rate budget, rejected-edit buffer, and epoch-wise slow/meta update make skill training stable while adding zero inference-time model calls at deployment. Across six benchmarks, seven target models, and three execution harnesses (direct chat, Codex, Claude Code), SkillOpt is best or tied on all 52 evaluated (model, benchmark, harness) cells and beats every per-cell competitor among human, one-shot LLM, Trace2Skill, TextGrad, GEPA, and EvoSkill skills. On GPT-5.5 it lifts the average no-skill accuracy by +23.5 points in direct chat, by +24.8 inside the Codex agentic loop, and by +19.1 inside Claude Code. Transfer experiments further show that optimized skill artifacts retain value when moved across model scales, between Codex and Claude Code execution environments, and to a nearby math benchmark without further optimization.
当前的智能体技能由人工编写、一次性生成，或通过控制宽松的自我修订来演化；这些方式都不像针对技能的深度学习优化器，也都无法在反馈下可靠地超越起点。我们认为，技能应被视为冻结智能体的外部状态来训练，并采用使权重空间优化可复现的同等严格方法。据我们所知，SkillOpt 是首个面向智能体技能的系统化、可控文本空间优化器：一个独立的优化器模型将带评分的展开过程转化为对单个技能文档的有界添加、删除或替换编辑，并且只有当编辑严格提升留出验证分数时才予以接受。文本学习率预算、被拒编辑缓冲区，以及按轮次进行的慢速更新与元更新，使技能训练保持稳定，同时在部署时不增加任何推理阶段模型调用。在六个基准、七个目标模型和三种执行框架（直接对话、Codex、Claude Code）上，SkillOpt 在全部 52 个被评测的“模型—基准—执行框架”组合中均取得最佳或并列最佳成绩，并在每个组合上击败人类技能、一次性大语言模型技能、Trace2Skill、TextGrad、GEPA 和 EvoSkill 等所有竞争方法。在 GPT-5.5 上，它相较于无技能设置将直接对话的平均准确率提高 23.5 个百分点，将 Codex 智能体循环中的准确率提高 24.8 个百分点，并将 Claude Code 中的准确率提高 19.1 个百分点。迁移实验进一步表明，优化后的技能产物跨模型规模迁移、在 Codex 与 Claude Code 执行环境之间迁移，以及无需进一步优化迁移到相近数学基准时，仍能保留其价值。
Mean Mode Screaming: Mean--Variance Split Residuals for 1000-Layer Diffusion Transformers
均值模态尖叫：用于千层扩散 Transformer 的均值-方差分离残差
Scaling Diffusion Transformers (DiTs) to hundreds of layers introduces a structural vulnerability: networks can enter a silent, mean-dominated collapse state that homogenizes token representations and suppresses centered variation. Through mechanistic auditing, we isolate the trigger event of this collapse as Mean Mode Screaming (MMS). MMS can occur even when training appears stable, with a mean-coherent backward shock on residual writers that opens deep residual branches and drives the network into a mean-dominated state. We show this behavior is driven by an exact decomposition of these gradients into mean-coherent and centered components, compounded by the structural suppression of attention-logit gradients through the null space of the Softmax Jacobian once values homogenize. To address this, we propose Mean-Variance Split (MV-Split) Residuals, which combine a separately gained centered residual update with a leaky trunk-mean replacement. On a 400-layer single-stream DiT, MV-Split prevents the divergent collapse that crashes the un-stabilized baseline; it tracks close to the baseline's pre-crash trajectory while remaining substantially better than token-isotropic gating methods such as LayerScale across the full schedule. Finally, we present a 1000-layer DiT as a scale-validation run at boundary scales, establishing that the architecture remains stably trainable at extreme depth.
将扩散 Transformer（DiT）扩展到数百层会引入一种结构性弱点：网络可能进入无声的均值主导坍缩状态，使 token 表示趋于同质并抑制中心化变化。通过机制审计，我们将这种坍缩的触发事件定位为均值模态尖叫（MMS）。即使训练表面上保持稳定，MMS 仍可能发生：残差写入模块上一次均值一致的反向冲击会打开深层残差分支，并将网络推入均值主导状态。我们表明，这种行为源于将这些梯度精确分解为均值一致分量与中心化分量；当数值趋于同质后，Softmax 雅可比矩阵的零空间还会在结构上抑制注意力 logit 梯度，从而进一步加剧这一问题。为解决该问题，我们提出均值-方差分离（MV-Split）残差，将具有独立增益的中心化残差更新与有泄漏的主干均值替换相结合。在 400 层单流 DiT 上，MV-Split 避免了导致未经稳定化基线崩溃的发散式坍缩；它的轨迹接近基线崩溃前的轨迹，并在整个训练计划中始终明显优于 LayerScale 等 token 各向同性门控方法。最后，我们使用一个 1000 层 DiT 在边界规模上进行规模验证，证明该架构在极端深度下仍可稳定训练。
Code as Agent Harness
https://github.com/YennNing/Awesome-Code-as-Agent-Harness-Papers
代码作为智能体运行框架
Recent large language models (LLMs) have demonstrated strong capabilities in understanding and generating code, from competitive programming to repository-level software engineering. In emerging agentic systems, code is no longer only a target output. It increasingly serves as an operational substrate for agent reasoning, acting, environment modeling, and execution-based verification. We frame this shift through the lens of agent harnesses and introduce code as agent harness: a unified view that centers code as the basis for agent infrastructure. To systematically study this perspective, we organize the survey around three connected layers. First, we study the harness interface, where code connects agents to reasoning, action, and environment modeling. Second, we examine harness mechanisms: planning, memory, and tool use for long-horizon execution, together with feedback-driven control and optimization that make harness reliable and adaptive. Third, we discuss scaling the harness from single-agent systems to multi-agent settings, where shared code artifacts support multi-agent coordination, review, and verification. Across these layers, we summarize representative methods and practical applications of code as agent harness, spanning coding assistants, GUI/OS automation, embodied agents, scientific discovery, personalization and recommendation, DevOps, and enterprise workflows. We further outline open challenges for harness engineering, including evaluation beyond final task success, verification under incomplete feedback, regression-free harness improvement, consistent shared state across multiple agents, human oversight for safety-critical actions, and extensions to multimodal environments. By centering code as the harness of agentic AI, this survey provides a unified roadmap toward executable, verifiable, and stateful AI agent systems.
近年来的大语言模型展现出强大的代码理解与生成能力，覆盖从竞赛编程到代码仓库级软件工程的多种任务。在新兴智能体系统中，代码不再只是目标输出，而是日益成为智能体进行推理、行动、环境建模和基于执行验证的运行基础。我们从智能体运行框架的视角描述这一转变，并提出“代码作为智能体运行框架”：一种以代码作为智能体基础设施核心的统一观点。为系统研究这一视角，我们围绕三个相互关联的层次组织综述。首先，我们研究运行框架接口，代码在其中将智能体连接到推理、行动和环境建模。其次，我们考察运行框架机制，包括面向长程执行的规划、记忆与工具使用，以及使运行框架可靠且自适应的反馈驱动控制与优化。第三，我们讨论如何将运行框架从单智能体系统扩展到多智能体设置，在这种设置中，共享代码产物支持多智能体协调、审查与验证。贯穿这些层次，我们总结代码作为智能体运行框架的代表性方法和实际应用，涵盖编程助手、GUI 与操作系统自动化、具身智能体、科学发现、个性化与推荐、DevOps 和企业工作流。我们还概述运行框架工程中的开放挑战，包括超越最终任务成功率的评测、不完整反馈下的验证、无回归的运行框架改进、多智能体间一致的共享状态、对安全关键操作的人工监督，以及向多模态环境扩展。通过将代码置于智能体 AI 运行框架的中心，本综述为构建可执行、可验证且有状态的 AI 智能体系统提供了一条统一路线图。
MinT: Managed Infrastructure for Training and Serving Millions of LLMs
Mind Lab
MinT：用于训练和服务数百万个大语言模型的托管基础设施
We present MindLab Toolkit (MinT), a managed infrastructure system for Low-Rank Adaptation (LoRA) post-training and online serving. MinT targets a setting where many trained policies are produced over a small number of expensive base-model deployments. Instead of materializing each policy as a merged full checkpoint, MinT keeps the base model resident and moves exported LoRA adapter revisions through rollout, update, export, evaluation, serving, and rollback, hiding distributed training, serving, scheduling, and data movement behind a service interface. MinT scales this path along three axes. Scale Up extends LoRA RL to frontier-scale dense and MoE architectures, including MLA and DSA attention paths, with training and serving validated beyond 1T total parameters. Scale Down moves only the exported LoRA adapter, which can be under 1% of base-model size in rank-1 settings; adapter-only handoff reduces the measured step by 18.3x on a 4B dense model and 2.85x on a 30B MoE, while concurrent multi-policy GRPO shortens wall time by 1.77x and 1.45x without raising peak memory. Scale Out separates durable policy addressability from CPU/GPU working sets: a tensor-parallel deployment supports 10^6-scale addressable catalogs (measured single-engine sweeps through 100K) and thousand-adapter active waves at cluster scale, with cold loading treated as scheduled service work and packed MoE LoRA tensors improving live engine loading by 8.5-8.7x. MinT thus manages million-scale LoRA policy catalogs while training and serving selected adapter revisions over shared 1T-class base models.
我们提出 MindLab Toolkit（MinT），一个用于低秩适配（LoRA）后训练与在线服务的托管基础设施系统。MinT 面向这样一种设置：在少量昂贵的基础模型部署之上生成大量训练后的策略。MinT 不会将每个策略实例化为合并后的完整检查点，而是让基础模型常驻，并使导出的 LoRA 适配器修订版本依次经过展开、更新、导出、评测、服务和回滚，同时将分布式训练、服务、调度和数据移动隐藏在服务接口之后。MinT 沿三个方向扩展这条路径。纵向扩展将 LoRA 强化学习扩展到前沿规模的稠密与混合专家架构，包括 MLA 和 DSA 注意力路径，且已在总参数量超过 1 万亿的模型上验证训练与服务。向下缩减仅移动导出的 LoRA 适配器；在秩为 1 的设置中，其大小可低于基础模型的 1%。仅交接适配器使 40 亿参数稠密模型上的实测步骤加速 18.3 倍，使 300 亿参数混合专家模型上的该步骤加速 2.85 倍；并发多策略 GRPO 在不提高峰值内存的情况下，分别将实际运行时间缩短 1.77 倍和 1.45 倍。横向扩展将持久策略寻址能力与 CPU/GPU 工作集分离：张量并行部署支持百万级可寻址目录，单引擎扫描实测达到 10 万项，并在集群规模上支持数千个适配器组成的活跃批次；冷加载被视为可调度的服务工作，打包后的混合专家 LoRA 张量则使在线引擎加载加速 8.5 至 8.7 倍。因此，MinT 能够管理百万级 LoRA 策略目录，同时在共享的万亿参数级基础模型上训练并服务选定的适配器修订版本。
Heterogeneous Scientific Foundation Model Collaboration
UIUC
异构科学基础模型协作
Agentic large language model systems have demonstrated strong capabilities. However, their reliance on language as the universal interface fundamentally limits their applicability to many real-world problems, especially in scientific domains where domain-specific foundation models have been developed to address specialized tasks beyond natural language. In this work, we introduce Eywa, a heterogeneous agentic framework designed to extend language-centric systems to a broader class of scientific foundation models. The key idea of Eywa is to augment domain-specific foundation models with a language-model-based reasoning interface, enabling language models to guide inference over non-linguistic data modalities. This design allows predictive foundation models, which are typically optimized for specialized data and tasks, to participate in higher-level reasoning and decision-making processes within agentic systems. Eywa can serve as a drop-in replacement for a single-agent pipeline (EywaAgent) or be integrated into existing multi-agent systems by replacing traditional agents with specialized agents (EywaMAS). We further investigate a planning-based orchestration framework in which a planner dynamically coordinates traditional agents and Eywa agents to solve complex tasks across heterogeneous data modalities (EywaOrchestra). We evaluate Eywa across a diverse set of scientific domains spanning physical, life, and social sciences. Experimental results demonstrate that Eywa improves performance on tasks involving structured and domain-specific data, while reducing reliance on language-based reasoning through effective collaboration with specialized foundation models.
智能体大语言模型系统已经展现出强大能力。然而，它们依赖语言作为通用接口，这从根本上限制了其在许多真实世界问题中的适用性，尤其是在科学领域；这些领域已经发展出面向特定领域的基础模型，以解决超越自然语言的专门任务。本文提出 Eywa，一个异构智能体框架，旨在将以语言为中心的系统扩展到更广泛的科学基础模型。Eywa 的核心思想是为特定领域基础模型增添基于语言模型的推理接口，使语言模型能够引导对非语言数据模态的推理。这种设计使通常针对专门数据与任务优化的预测式基础模型能够参与智能体系统中的高层推理和决策过程。Eywa 可以直接替换单智能体流水线中的原有组件（EywaAgent），也可以通过用专门智能体替代传统智能体而集成到现有多智能体系统中（EywaMAS）。我们还研究了一种基于规划的编排框架，其中规划器动态协调传统智能体和 Eywa 智能体，以解决横跨异构数据模态的复杂任务（EywaOrchestra）。我们在物理科学、生命科学和社会科学等多种科学领域评测 Eywa。实验结果表明，Eywa 提升了涉及结构化数据和特定领域数据的任务性能，并通过与专门基础模型有效协作，降低了对基于语言推理的依赖。
DelTA: Discriminative Token Credit Assignment for Reinforcement Learning from Verifiable Rewards
DelTA：面向可验证奖励强化学习的判别式 Token 信用分配
Reinforcement learning from verifiable rewards (RLVR) has emerged as a central technique for improving the reasoning capabilities of large language models. Despite its effectiveness, how response-level rewards translate into token-level probability changes remains poorly understood. We introduce a discriminator view of RLVR updates, showing that the policy-gradient update direction implicitly acts as a linear discriminator over token-gradient vectors and thereby determines which token probabilities are increased or decreased during learning. Under standard sequence-level RLVR, this discriminator is constructed from positive- and negative-side centroids formed by advantage-weighted averaging of token-gradient vectors. However, such centroid construction can be dominated by shared high-frequency patterns, such as formatting tokens, diluting sparse yet discriminative directions that better distinguish high-reward responses from low-reward ones. To address this limitation, we propose DelTA, a discriminative token credit assignment method that estimates token coefficients to amplify side-specific token-gradient directions and downweight shared or weakly discriminative ones. These coefficients reweight a self-normalized RLVR surrogate, making the effective side-wise centroids more contrastive and thereby reshaping the RLVR update direction. On seven mathematical benchmarks, DelTA outperforms the strongest same-scale baselines by 3.26 and 2.62 average points on Qwen3-8B-Base and Qwen3-14B-Base, respectively. Additional results on code generation, a different backbone, and out-of-domain evaluations further demonstrate the generalization ability of DelTA.
可验证奖励强化学习（RLVR）已经成为提升大语言模型推理能力的核心技术。尽管它行之有效，但回答级奖励如何转化为 token 级概率变化仍缺乏充分理解。我们提出一种看待 RLVR 更新的判别器视角，表明策略梯度的更新方向会隐式充当 token 梯度向量上的线性判别器，从而决定学习过程中哪些 token 概率升高或降低。在标准序列级 RLVR 中，该判别器由正侧与负侧质心构成，这些质心通过对 token 梯度向量进行优势加权平均得到。然而，这种质心构造可能被格式 token 等共享高频模式主导，从而稀释那些稀疏但更具判别性、能够更好地区分高奖励回答与低奖励回答的方向。为解决这一局限，我们提出 DelTA，一种判别式 token 信用分配方法。它估计 token 系数，以增强特定侧的 token 梯度方向，并降低共享方向或判别性较弱方向的权重。这些系数对自归一化 RLVR 替代目标重新加权，使有效的两侧质心更具对比性，进而重塑 RLVR 的更新方向。在七个数学基准上，DelTA 在 Qwen3-8B-Base 和 Qwen3-14B-Base 上分别比同等规模的最强基线平均高出 3.26 和 2.62 个百分点。代码生成、不同骨干模型和分布外评测上的额外结果进一步证明了 DelTA 的泛化能力。