论文

本文提出 When2Think，一种基于 RLVR 的后训练框架，用于实例自适应计算分配，既无需学习奖励模型，也无需学习 critic；离线参考缓存机制避免了策略更新阶段对参考模型的在线查询。This work proposes When2Think, an RLVR-based post-training framework for instance-adaptive computation allocation that requires neither a learned reward model nor a learned critic, and offline reference caching avoids online reference-model queries during policy updates.

提出 Calibrated On-Policy Distillation：通过正、负特权干预估计教师的自偏离区间，并仅保留超出该区间的部分以校准原始的 teacher–student 差异。Calibrated On-Policy Distillation is introduced, which estimates the teacher's self-deviation region through positive and negative privileged interventions and calibrates the original teacher--student discrepancy by retaining only the component that lies beyond this region.

本文使用 Fast Iterative Shrinkage-Thresholding Algorithm 展开 CSC 优化，并将稀疏系数视为可微变量与网络参数联合学习；同时提出一种 label-free 的后训练策略，在固定主网络参数的情况下，根据被损坏输入自适应调整压缩强度。This work unfolds the CSC optimization with the Fast Iterative Shrinkage-Thresholding Algorithm and treats the sparsity coefficient as a differentiable variable jointly learned with the network parameters and introduces a label-free post-training strategy that adjusts the compression strength for corrupted inputs with the main network parameters fixed.

RefineEdit 是一个基于 GRN 的 training-free prompt-to-prompt 图像编辑框架，将 bit routing 与两种稳定机制（adaptive spatial freezing 与 finite bit locking）相结合，使编辑证据能够随图像演化而被修正。RefineEdit is a training-free prompt-to-prompt image editing framework built on the GRN that combines bit routing with two stabilization mechanisms: adaptive spatial freezing and finite bit locking, allowing editing evidence to be revised as the image evolves.

提出 PARTS（Policy Adaptation with RL on Targeted Subtasks），一种真实世界子任务强化学习框架，能够将训练集中于瓶颈环节，同时以最小的人工干预推进训练 rollout。PARTS (Policy Adaptation with RL on Targeted Subtasks), a real-world subtask RL framework that concentrates practice at bottlenecks while allowing training rollouts to proceed with minimal human intervention, is presented.

证明了对通用语言模型适配教育任务（涵盖问题求解、课程理解与教学辅助）的精选、能力均衡的监督数据的价值。The value of curated, capability-balanced supervision for adapting general language models to educational tasks spanning problem solving, curriculum understanding, and instructional support is demonstrated.

提出 Calibrated Clipping，一种动态方法，通过匹配下界裁剪分位数并相应重平衡上界，将 FP8 裁剪边界与高精度 BF16 分布对齐，消除熵激增并恢复与 BF16 基线可比的性能。Calibrated Clipping is proposed, a dynamic method that aligns the FP8 clipping bounds with high-precision BF16 distributions by matching the lower-bound clipping quantile and rebalancing the upper bound accordingly, which eliminates entropy surges and restores performance comparable to the BF16 baseline.

提出 StableVQ，重新审视每个模块的合适学习目标，解决各模块独立训练以承担各自角色时产生的问题；其基于共享投影 codebook 构建，轻量且不引入可学习参数。StableVQ is proposed, which revisits the proper learning objective of each module and resolves the problems that arise when each is trained to fulfill its own role independently, and Built on top of shared-projection codebooks, is lightweight and introduces no learnable parameters.

本文提出了 SEGMENT-SNAP，通过 part-handle coupling 融合几何与语义证据，在 Articulate3D Challenge 中获得第一名。This work presents SEGMENT-SNAP, which combines geometric and semantic evidence through part-handle coupling and achieved first place in the Articulate3D Challenge.

论文表明，在该高维空间中使用 flow matching 的标准速度预测要求模型拟合低维信号流形之外的正交噪声方向，导致优化效率低下，由此提出改用清晰数据参数化（$\boldsymbol{x}_{0}$-prediction），使学习聚焦于底层信号流形。It is shown that using the standard velocity prediction in flow matching in this high-dimensional space requires the model to fit orthogonal noise directions outside the low-dimensional signal manifold, making optimization inefficient, and motivated using the clean data parameterization ($\boldsymbol{x}_{0}$-prediction) instead, which focuses learning on the underlying signal manifold.

PackLab 是一个用于开发、训练与评估闭环机器人装箱 MLLM 的综合框架，在不同物体集合与容器配置下均优于传统装箱启发式方法、经典强化学习方法以及通用 MLLM，展现了 MLLM 在长时任务机器人装箱中的潜力。PackLab is a comprehensive framework for developing, training, and evaluating MLLMs for closed-loop robotic bin packing that outperforms conventional packing heuristics, traditional reinforcement learning methods, and general-purpose MLLMs across object sets and container configurations, highlighting the potential of MLLMs for long-horizon robotic packing.

本文提出一个基于原理的、无需训练的框架，依次优化跨层权重配对与共享字典分解，并识别结构兼容的投影、学习一种能更好保留各层独立校准几何的共享表征。This work introduces a principled, training-free framework that sequentially optimizes cross-layer weight pairings and shared-dictionary factorizations, and identifies structurally compatible projections and learns a shared representation that better preserves each layer's distinct calibration geometry.

主要发现是：多样且高质量的 SFT 奠定坚实的能力下限，难度过滤将 RL 提示维持在有效的学习区间内，而奖励可靠性为阶段排序提供了实用原则。The main findings are that diverse, high-quality SFT establishes a strong capability floor and difficulty filtering keeps RL prompts within a productive learning range, and reward reliability provides a practical principle for ordering stages.

本文识别出两种强化水印证据对载体图像依赖性、抑制残余可迁移性的机制，并提出 CoverLock，一种即插即用策略，可在不重新设计架构的前提下强化现有水印系统的图像依赖性。This work identifies two mechanisms that strengthen the dependence of watermark evidence on the cover image, thereby suppressing the residual transferability, and introduces CoverLock, a plug-and-play strategy for existing watermarking systems that strengthens such image dependence without architectural redesign.

Nereus 是一种成本感知的运行时，将 RL 后训练任务适配为高效执行计划，并基于内存可行的全局计划执行状态转移，使用与运行任务校准的成本模型来接纳转移。Nereus is a cost-aware runtime that adapts RL post-training jobs into efficient execution plans and executes a transition using a memory-feasible global plan and admits the transition using a cost model calibrated against the running job.

注意力往往集中在上下文中一小部分 token 上，但每个 query 关注的关键子集各不相同。为利用这种动态结构，我们提出 SANTA++，一种免训练的随机注意力方法，通过代表性 key 进行内存高效的选择，无需扫描整个 KV cache。缓存的 key 被组织成若干 team，query 对每个 team 中的代表性 key 打分以决定采样哪些 team。我们在采样得到的 team 内计算精确的注意力分数，并通过其采样概率的倒数对各 team 的贡献进行重新加权。Attention often concentrates on a small subset of tokens in the context, but which subset matters changes from one query to the next. To exploit this changing structure, we introduce SANTA++, a training-free stochastic attention method that uses representative keys for memory-efficient selection without scanning the entire key-value (KV) cache. Cached keys are organized into teams, and the query scores one representative from each team to decide which teams to sample. We compute exact attention scores within the sampled teams and reweight each team's contribution by the inverse of its inclusio

结果表明，推理路径多样性可作为筛选 SFT 数据的实用准则，能更好地为 RL 准备模型，并据此提出一种轻量级、基于规则的指纹方法用于筛选。These results identify reasoning-route diversity as a practical criterion for selecting SFT data that better prepares models for RL, and propose a lightweight, rule-based fingerprint to select for it.

提出 Chinese-Jev，一个 System One 模型，通过统一的数据处理与训练流水线弥合预训练分布与下游中文场景间的差距，在各专业领域上达到 Jev 平均准确率的 92%。Chinese-Jev is introduced, a System One model that addresses the gap between the pre-training distribution and downstream Chinese-language scenarios through a unified data processing and training pipeline, and achieves 92% of Jev's average accuracy across specialized domains.

本文推导了对应的反向传播规则（包括校准项的导数），并在模型、数据、优化器均一致的预训练实验中对不同选择进行了比较。This work derives the corresponding backward rules, including calibration derivatives, and compares these choices in pretraining experiments matched on model, data, and optimizer, and compares these choices in pretraining experiments matched on model, data, and optimizer.

研究发现，早期 OPD 训练动态均呈现一种规律的 *useful-transfer* 区间，其中留出准确率（即 *gold score*, $G$）随 $d=\sqrt{\mathrm{KL}(\pi_\theta \Vert \pi_{\mathrm{ref}})}$（学生初始化在 token 级反向 KL 散度的平方根）近似线性上升。It is found that early OPD training dynamics uniformly exhibit a regular *useful-transfer* regime, in which held-out accuracy (the *gold score*, $G$) rises approximately linearly in $d=\sqrt{\mathrm{KL}(\pi_\theta \Vert \pi_{\mathrm{ref}})}$, the square root of token-level reverse KL divergence from the student initialization.

本文提出 ATLAS（Aligned Transport of Latent Structure），一种在显式保持关系几何结构的同时校准全局潜空间分布的训练目标，通过一维 Wasserstein-2 传输进行 Wasserstein 嵌入匹配来校准其边缘分布。This work introduces Aligned Transport of Latent Structure (ATLAS), a training objective that explicitly preserves relational geometry while calibrating the global latent distribution and uses Wasserstein embedding matching to calibrate its marginal through one-dimensional Wasserstein-2 transport.

基于 Rec 的检索在所有划分上都提升了回访一致性；当上下文跨度足以覆盖每次返回的首次访问时，可再带来 24% 到 30% 的提升，但会牺牲一定图像质量；超过该跨度后，继续增加长度不再带来收益。Rec retrieval raises revisit consistency on every split, and a span long enough to reach the first visit of each return adds a further 24% to 30%, at some cost in image quality; beyond that span, more length no longer helps.

在统计物理中，多元硬核模型描述一个粒子系统，每个粒子拥有各自的逸度。用图论语言表述，该模型的配分函数对应多元独立多项式，即独立多项式的多重仿射推广，定义为 $Z_G(λ_1,\dots,λ_n) := \sum_{I\in\mathcal{I}(G)} \prod_{v\in I}λ_v$，其中 $\mathcal{I}(G)$ 表示 $[n]:=\{1,2,\dots,n\}$ 上图 $G$ 的所有独立集。我们证明对于 $[n]$ 上的每个简单图 $G$ 以及 $λ_1,\dots,λ_n\geq 0$，\[ Z_G(λ_1,\dots,In statistical physics, the multivariate hard-core model describes a system of particles, each of which receives its own fugacity. In graph-theoretic language, the partition function of the model translates to the multivariate independence polynomial, i.e., the multiaffine generalisation of the independence polynomial, defined by $Z_G(λ_1,\dots,λ_n) := \sum_{I\in\mathcal{I}(G)} \prod_{v\in I}λ_v$, where $\mathcal{I}(G)$ denotes the set of all independent sets in a graph $G$ on $[n]:=\{1,2,\dots,n\}$. We prove that for every simple graph $G$ on $[n]$ and $λ_1,\dots,λ_n\geq 0$, \[ Z_G(λ_1,\dots,

预训练 Transformer 仅利用其深度的一小部分来跟踪上下文中的引用。十三个基础模型仅能可靠地跟随 1.4–3.6 行，额外的预训练循环收益甚微。在一个早期层上训练的 rank-8 LoRA 在所有模型权重冻结的情况下扩展了这一计算能力。Qwen3-8B 在 24 行链上的精确准确率从 15.5% 提升至 99%；更长训练的 LoRA 可达 50 行。Ouro-1.4B 经过四轮循环达到 60 行，八轮后至少达到 160 行。该 LoRA 启动了一场接力：程序行通过中间层的一段短距离传递其链身份。冻结的 head 逐层读取渐进式进展信号……Pretrained transformers use little of their depth to follow references in context. Thirteen base models reliably follow only 1.4-3.6 lines, and extra pretrained loops add little. A task-trained rank-8 LoRA at one early layer extends this computation with all model weights frozen. Qwen3-8B improves from 15.5% to 99% exact accuracy on 24-line chains; a longer-trained LoRA reaches 50 lines. Ouro-1.4B reaches 60 lines after four loops and at least 160 after eight. The LoRA starts a relay: program lines pass on their chain identity through a short range of middle layers. Frozen heads read progressi

结果将控制器所学到的行为范围与该范围内请求的准确性分离开来，尽管连贯性下限并非对每种特性都成立。The results separate the behavioral range learned by a controller from the accuracy of requests within that range, although the coherence floor does not hold for every trait there.

本文用三种以不同方式施加长度压力的方法（固定生成预算、逐样本长度目标、组相对长度奖励）对多种模型进行微调，发现它们对忠实性和可监控性具有不同的影响。This work fine-tune a variety of models with three methods that apply length pressure differently, namely a fixed generation budget, a per-example length target, and a group-relative length reward, and finds that it affects faithfulness and monitorability differently.

JEPA-TTT 在测试时间内持续适配预训练动作条件 Joint-Embedding Predictive Architecture 世界模型的潜在动力学预测器，平均将自回归潜在预测误差降低 83%，并将规划性能相较冻结的 JEPA 世界模型提升 153%。JEPA-TTT, which adapts the latent dynamics predictor of a pretrained action-conditioned Joint-Embedding Predictive Architecture world model throughout test time, reduces autoregressive latent prediction error by 83% on average and improves planning performance by 153% over the frozen JEPA world model.

结果表明，面向框架的预训练、经过验证的 SFT 以及显式的能力保留评估，分别针对领域特定可执行代码生成中不同的失效模式。The results show that framework-oriented pretraining, validated SFT, and explicit capability-retention evaluation address distinct failure modes in domain-specific executable code generation.

表格基础模型通过以标注训练样本为条件进行上下文学习 (ICL) 来预测。与分离训练与推理的传统模型不同，这些模型必须在每次前向传播中处理所有训练样本，导致每次预测成本高昂。限制训练样本数量可降低成本，但会显著降低性能。我们不丢弃上下文，而是提出激活对齐 (activation alignment)，利用完整上下文来教会模型在仅看到子集时如何行为。这是通过训练一个……Tabular foundation models perform in-context learning (ICL) by conditioning predictions on labeled training examples provided as context. Unlike traditional models that separate training from inference, these models must process all training examples in every forward pass, making each prediction expensive. Restricting the number of training examples reduces this cost but substantially degrades performance. Instead of discarding context, we propose activation alignment, a method that leverages the full context to teach a model how to behave when seeing only a subset. This is achieved by trainin

提出 LLM-as-Jev，一种保留架构的框架，直接从带括号数字标识符的下一 token 概率中提取校准决策，并同时给出无需训练的推理方案与一种微调目标——通过树分解的列表式损失优化候选选择，并使用 KL 散度惩罚将辅助预测锚定到基模型。This work presents LLM-as-Jev, an architecture-preserving framework that extracts calibrated decisions directly from next-token probabilities over bracketed numeric identifiers, and provides both a training-free inference recipe and a fine-tuning objective that optimizes candidate selection via a tree-factorized listwise loss while anchoring auxiliary predictions to the base model using KL divergence penalties.

强化学习（RL）是推动大型基础模型走向自我改进的核心训练范式。本报告介绍 MiMo-V2.6 系列，这是一个通过扩展 RL 算力来推动模型智能边界的 omni-modal 模型族。在 RL 之前，我们在广泛的跨模态语料上进行 mid-training 以提供充足的探索空间，并基于预训练的 hybrid-SWA 架构构建稳健的基础设施以支持后续规模化。我们沿三个维度扩展 RL 算力：(1) 更大的 batch 和更高的吞吐量，采用异步训练来持续消费Reinforcement learning (RL) is the central training paradigm for advancing large foundation models towards self-improvement. This report introduces the MiMo-V2.6 series, an omni-modal family that pushes the frontier of model intelligence by scaling RL compute. Prior to RL, we conduct mid-training on a broad multimodal corpus to provide ample exploration space, and build a solid infrastructure on the pretrained hybrid-SWA architecture to support subsequent scale-up. We scale RL compute along three dimensions: (1) larger batches and higher throughput, with an asynchronous training that consumes

CONFLUX 是一个面向胸部 CT 的潜在扩散模型：由 3D 变分自编码器压缩每个体数据，整流流 Transformer 在潜在空间中生成，并以分类器从生成体中恢复所请求病征的可靠性作为奖励。CONFLUX, a latent diffusion model for chest computed tomography (CT): a 3D variational autoencoder compresses each volume, and a rectified-flow transformer generates in the latent space, that rewards how reliably a classifier recovers the requested findings from each generated volume.

首个面向预训练胸部 X 光报告生成器的无需训练 best-of-N 采样方案，显式建模"既往-当前-过渡"纵向先验，全面优于随机选择。This work presents the first training-free best-of-N sampling scheme for pre-trained chest X-ray report generators that is explicitly aware of this longitudinal prior to current transition, and outperforms random selection across the board.

本文提出了 Graph-PRefLexOR，这是一族基于图结构的推理模型，使用 Group Relative Policy Optimization 进行微调，将推理过程组织为显式阶段，分别用于机理探索、图构建、模式提取和假设合成，确立了面向图结构的强化学习作为通向可解释 AI 系统的路径，可应用于材料设计及其他科学领域的科学假设生成。Graph-PRefLexOR is developed, a family of graph-native reasoning models fine-tuned with Group Relative Policy Optimization to organize reasoning into explicit phases for mechanism exploration, graph construction, pattern extraction, and hypothesis synthesis, establishing graph-native reinforcement learning as a pathway toward interpretable AI systems for scientific hypothesis generation in materials design and other scientific applications.

本文提出 Monotonic Inference Policy Update (MIPU)，一种两步式 LLM 强化学习框架：构建采样器引用的候选更新，并使用推理侧差距代理选择性接受同步候选；实验表明 MIPU 提升了平均推理性能与训练稳定性。Monotonic Inference Policy Update (MIPU) is introduced, a two-step LLM RL framework that constructs sampler-referenced candidate updates and selectively accepts synchronized candidates using an inference-side gap proxy, and experiments show that MIPU improves average reasoning performance and training stability.

本文提出 MultiDepth-3k (MD-3k)，一个用于衡量深度层偏好与多层空间关系准确率 (ML-SRA) 的稀疏双层序数基准；领先的深度基础模型在标准 RGB 输入下表现出不同的层偏好，表明同一分层几何可在不同模型中被差异化地解析。MultiDepth-3k (MD-3k), a sparse two-layer ordinal benchmark for measuring depth-layer preference and multi-layer spatial relationship accuracy (ML-SRA), is introduced and leading depth foundation models exhibit diverse layer preferences under standard RGB input, showing that the same layered geometry can be resolved differently across models.