HDATF
简体中文
开始交流
菜单

论文档案库

按编号从大到小列出全部收录论文及摘要。

收录论文 1,430资料更新日期 Research Monitoring基准测试
  1. 381AI / GeneralarXiv:2606.08432

    Trajectory-Refined Distillation

    • It is the first to formalize the prefix-failure problem in OPD, showing that token-level improvements cannot solve it at the root.
    • TRD performs distillation after refining trajectories within the on-policy support region through trajectory-level refinement.
    • It can be applied to both OPD and OPSD self-distillation and yields consistent gains across the Qwen3 model family.
    • On AMOBench, Qwen3-8B shows roughly 50 percent relative improvement in Pass@16, and the refined trajectories are compressed by about 9 times.
    • This trajectory-level approach outperforms prior token-level loss-improvement methods such as clipping, reweighting, and top-K selection.
    原文摘要PDF
  2. 380LLMs / NLParXiv:2606.09498

    Self-Harness: Harnesses That Improve Themselves

    • It proposes Self-Harness, a paradigm in which the agent improves its own harness through weakness mining, harness proposal, and proposal validation.
    • Across three models, MiniMax M2.5, Qwen3.5-35B-A3B, and GLM-5, the held-out pass rates improve from 40.5% to 61.9%, from 23.8% to 38.1%, and from 42.9% to 57.1%, respectively.
    • Different harness changes are produced for different models, so the updates are model-specific rather than generic instruction additions.
    • A regression-test-based acceptance rule prevents overfitting, and the edits remain small and auditable.
    • The work is directly related to existing harness systems such as SemaClaw and OpenClaw.
    • It is a meaningful step forward in the self-improvement research line for agents.
    原文摘要PDF
  3. 379Computer VisionarXiv:2606.09056

    MilliVid: Hierarchical Latents for Long-Range Consistency in Video Generation

    • It uses a hierarchical autoencoder design with different token counts per frame at different levels.
    • The coarsest level, with 4 tokens per frame, captures scene structure and object permanence.
    • Coarse-to-fine rollout maximizes temporal coverage within a fixed sequence length.
    • Learned hierarchical tokenization provides better compression quality than simple downsampling.
    • It trains the model to predict far-future frames, encouraging long-range context dependence.
    • It shows a significant improvement in consistency over FramePack on long Minecraft videos.
    原文摘要PDF
  4. 378AI / GeneralarXiv:2606.06741

    OpenSkill: Open-World Self-Evolution for LLM Agents

    • Open-world self-evolution is defined as the process of building skills and verification signals at the same time from task prompts alone.
    • In the open world, grounding knowledge and verification anchors are acquired independently.
    • Self-built virtual tasks are used for skill refinement, with no target supervision at all.
    • On SkillsBench, it improves by 8.9 percent over the previous strongest baseline.
    • The skills transfer across models, so the same effect applies even to weaker models.
    • The self-built verifier covers 88.9 percent of the intent represented in the ground truth.
    原文摘要PDF
  5. 377Software EngineeringarXiv:2606.07297

    SWE-Explore: Benchmarking How Coding Agents Explore Repositories

    • A new benchmark that evaluates repository exploration independently from patch generation.
    • It derives trajectory-grounded, line-level ground truth from successful agent trajectories.
    • It covers diverse coding settings with 848 issues, 10 languages, and 203 repositories.
    • Agent exploration forms a clear top tier compared with classical retrieval.
    • File-level hit rates are strong, but core line recall remains only 0.14 to 0.19, making it the bottleneck.
    • Missing key context is more damaging to patch success than noisy context.
    原文摘要PDF
  6. 376AI / GeneralarXiv:2606.08702

    ConMem: Structured Memory-Guided Adaptation in Training-Free Multi-Agent Systems

    • It distills past trajectories into typed signed memory cards to create reusable strategic units.
    • It organizes the cards as typed graphs with supports, satisfies, conflicts, and constrains relations.
    • It uses needs-aware retrieval to search for cards that match the current task.
    • It resolves conflicts and recovers dependencies through the relation graph to produce consistent guidance.
    • Context-budget control prunes more than 50% of candidate cards and reduces planning overhead by 80%.
    • It is training-free and does not require changes to the host model weights or orchestration.
    原文摘要PDF
  7. 375LLMs / NLParXiv:2606.07502

    Your UnEmbedding Matrix is Secretly a Feature Lens for Text Embeddings

    • It discovers representation collapse in LLM embeddings, where they align with high-frequency meaningless tokens.
    • It identifies the edge-spectrum subspace of the unembedding matrix as the cause.
    • EmbedFilter removes that subspace with a simple linear transform and does not require additional training.
    • It improves MTEB performance by up to 14.1% and works across different LLM backbones.
    • It also saves storage and speeds up retrieval through distance-preserving dimensionality reduction.
    • It outperforms existing correction methods such as BERT-whitening in unsupervised settings.
    原文摘要PDF
  8. 374RoboticsarXiv:2606.09811

    AHA-WAM:Asynchronous Horizon-Adaptive World-Action Modeling with Observation-Guided Context Routing

    • The design separates Video DiT as a low-frequency planner and Action DiT as a high-frequency executor through an asynchronous decomposition.
    • Rolling KV memory preserves the planner's temporal context over time.
    • OVCR dynamically routes and updates the planner context based on the latest observations.
    • Horizon-adaptive offset training learns arbitrary phase relationships between planner and executor.
    • The method reaches SOTA-level performance, with 92.80 percent on RoboTwin and 78.3 percent success in the real world.
    • It achieves control frequencies up to 56.9 Hz, which is a 10.82x speedup over Fast-WAM.
    原文摘要PDF
  9. 373Machine LearningarXiv:2606.09079

    FlashMemory-DeepSeek-V4: Lightning Index Ultra-Long Context via Lookahead Sparse Attention

    • LSA is a paradigm that predicts future context demand and loads only the necessary KV chunks onto the GPU.
    • It decouples the indexer from the backbone, making it possible to train on a single H20 GPU in one hour.
    • It replaces fixed Top-k with dynamic chunk recall based on a sigmoid threshold.
    • It improves accuracy by 0.6% while using only 13.5% KV cache on average.
    • At 500K length, it cuts memory use by 90% and still improves accuracy by 1.9%.
    • It has a clear limitation: performance drops sharply on the dense-search MRCR benchmark, from 76% to 48%.
    原文摘要PDF
  10. 372AI / GeneralarXiv:2606.09669

    SpatialWorld: Benchmarking Interactive Spatial Reasoning of Multimodal Agents in Real-World Tasks

    • Design a simulator-agnostic evaluation protocol that integrates eight simulators.
    • Build a dataset of 760 tasks across six domains, including household chores, travel, social cooperation, and digital games.
    • Provide an MLLM-native interface with first-person vision-only observations and text-based actions.
    • Space reasoning remains a major challenge, with GPT-5 at 17.4% and Qwen-3.5 at 14.1%.
    • Domain-specific leaders differ: GPT-5 leads on household chores, while Gemini-3.1-Pro leads on games.
    • It finds a mismatch between task success rate and execution efficiency.
    原文摘要PDF
  11. 371Computer VisionarXiv:2606.09828

    Latent Spatial Memory for Video World Models

    • We build 3D spatial memory directly in latent space, eliminating the cost of round-tripping through pixel space.
    • Depth-guided back-projection loads latent tokens into 3D world space.
    • Occlusion-aware latent-resolution projection is used to obtain the target view.
    • We exclude dynamic objects and the sky from memory updates to improve long-term stability.
    • Compared with an RGB cache, the system runs 10.57 times faster end to end and uses 55 times less GPU memory.
    • Two-stage training enables stable convergence of the side branch and LoRA.
    原文摘要PDF
  12. 370RoboticsarXiv:2606.06556

    Robots Need More than VLA and World Models

    • It redefines VLA models as one layer in the physical intelligence stack and argues that upstream and downstream grounding mechanisms are essential.
    • It advocates a shift to grounding-centered pipelines that convert unstructured physical experience into robot-usable supervision.
    • It proposes four essential interfaces: a physical data engine, embodiment retargeting, physics-based world models, and task-conditioned reward grounding.
    • It presents a vision for self-improving systems that use success, failure, and human correction as structured supervision in the deployment loop.
    • It revisits state-of-the-art research in robot-native data, video-based methods, simulation, and world models from the perspective of grounding bottlenecks.
    原文摘要PDF
  13. 369Computer VisionarXiv:2606.07433

    Watch, Remember, Reason: Human-View Video Understanding with MLLMs

    • It formalizes video understanding as the functional process of observing, remembering, and reasoning, and proposes an integrated formulation for perceptual representations, memory state, reasoning traces, and final prediction.
    • It systematically organizes the viewing domain into four subareas: fine-grained perception, holistic perception, audiovisual perception, and efficient processing.
    • In memory, it distinguishes offline memory from streaming memory to analyze the tension between redundancy and evidence sparsity in long videos.
    • In reasoning, it integrates and classifies different paradigms such as agentic versus non-agentic and text-only versus video-involved reasoning.
    • It covers five application domains, including egocentric, sports, instructional, medical, and narrative, along with major training datasets and evaluation benchmarks.
    原文摘要PDF
  14. 368Machine LearningarXiv:2606.06624

    Principles and Practice of Deep Representation Learning: or a Mathematical Theory of Memory

    • It presents data compression as a unifying principle behind all representation learning methods, including PCA, denoising, score matching, and information gain maximization.
    • It mathematically interprets ResNet, CNN, and Transformer architectures as iterative optimization algorithms and clarifies the principles behind architecture design.
    • It proves that the encoder-decoder structure is a necessary design for ensuring the accuracy and consistency of representations through a closed-loop transcription framework.
    • In v2.0, it adds the theoretical justification for contrastive learning and DINO, derives a causal CRATE transformer, and expands the relationship between distribution learning and representation learning.
    • It demonstrates the practical value of these theoretical principles step by step in large-scale real-world applications involving visual, motion, and text data.
    原文摘要PDF
  15. 367Computer VisionarXiv:2606.06476

    Thinking with Imagination: Agentic Visual Spatial Reasoning with World Simulators

    • It formulates spatial reasoning as an interactive evidence-gathering problem, so the VLM actively requests and uses imagined visual evidence.
    • It applies view-consistency tuning to Astra-WM to address the lack of spatial consistency in existing image generation models.
    • A two-stage RL curriculum teaches valid simulator calls in stage 1 and encourages selective imagination in stage 2.
    • It shows that simply connecting a simulator to Qwen3-VL degrades performance, but meaningful gains appear after RL training.
    • It shows that useful imagination requires both a high-quality world simulator and an agent policy that decides when, where, and how to imagine.
    原文摘要PDF
  16. 366Computer VisionarXiv:2606.07100

    LARA: Latent Action Representation Alignment for Vision-Language-Action Models

    • We overcome the limitations of existing separate training pipelines by jointly optimizing LAM and VLA with a bidirectional representation-alignment loss.
    • LAM is grounded in real action trajectories, which suppresses learning spurious visual variations such as background and lighting.
    • VLA is regularized by LAM's forward-dynamics prediction, reducing hallucinated trajectories that are kinematically feasible but functionally wrong.
    • We find that inserting alignment at DiT layer L-2 is optimal, and that alignment in deeper layers is effective.
    • We achieve consistent performance gains on three simulation benchmarks and in a real robot manipulation environment.
    原文摘要PDF
  17. 365LLMs / NLParXiv:2606.07229

    MMAE: A Massive Multitask Audio Editing Benchmark

    • It establishes a unified taxonomy for audio editing for the first time, spanning seven modalities, six complexity levels, and eight operation types.
    • It introduces an evaluation paradigm that decomposes free-form editing tasks into a verifiable multidimensional rubric, which enables objective and interpretable assessment.
    • It builds the benchmark from 2,000 high-quality samples and 17,741 evaluation rubrics selected through a human-agent collaboration pipeline.
    • Evaluation of five state-of-the-art models shows an EMR below 5 percent and 0 percent on complex mixed modalities, which exposes the fundamental limitations of current systems.
    • It confirms that introducing an external agent planner does not consistently help and that both understanding and generation remain bottlenecks.
    原文摘要PDF
  18. 364AI / GeneralarXiv:2606.07080

    dots.tts Technical Report

    • It trains a high-resolution AudioVAE that encodes 48 kHz speech into a 25 Hz, 128-dimensional latent space and adds semantic structure through WavLM alignment loss.
    • It uses a three-module split of semantic encoder, LLM, and AR flow-matching head to reduce long-range error accumulation.
    • It introduces SOAR self-corrective alignment, which exposes the flow-matching head to its own inference errors without a reward model.
    • CFG-aware MeanFlow distillation enables 2 to 4 NFE inference steps and achieves 0.231 RTF and 54 ms first-packet latency on H800.
    • It reports open-source SOTA on Seed-TTS-Eval with WER 2.92 and SIM 79.2, and releases the full system under Apache 2.0.
    原文摘要PDF
  19. 363Machine LearningarXiv:2606.07082

    On the Geometry of On-Policy Distillation

    • OPD lies in a relaxed off-principal regime between SFT, which aligns principal components, and RLVR, which localizes non-principal components.
    • The authors discover subspace locking, in which updates lock into a narrow low-dimensional channel soon after training begins.
    • Under a rank-16 projection constraint, OPD preserves performance while SFT degrades sharply, which shows that the locked subspace is functionally sufficient.
    • Token sparsification and off-policy rollouts preserve the rank trajectory, but objective mixing changes it, which confirms that objective composition is the key control axis.
    • Controlled experiments based on Qwen3-8B use multidimensional diagnostic metrics such as update sparsity, spectral drift, and principal-component rotation.
    原文摘要PDF
  20. 362AI / GeneralarXiv:2606.05661

    Continual Learning Bench: Evaluating Frontier AI Systems in Real-World Stateful Environments

    • It is a set of high-quality, expert-validated tasks from six real-world domains, and each task contains latent structure that cannot be solved through pretraining alone.
    • It proposes an evaluation methodology that introduces a gain metric to separate base model capability from experience-based learning.
    • Simple in-context learning (ICL) outperforms dedicated memory systems such as Mem0 and ACE on most tasks.
    • Even the best system reaches only 25.4% normalized gain, suggesting that the continual learning ability of current frontier models is still limited.
    • Memory modules are frequently observed to hurt performance by producing false generalizations and stale beliefs.
    • In terms of cost efficiency, ICL is also superior, and expensive systems fail to deliver proportional gains.
    原文摘要PDF
  21. 361Information RetrievalarXiv:2606.06260

    OneReason Technical Report

    • The paper analyzes the cause of the problem in generative recommendation, where thinking mode fails to outperform non-thinking mode, along two axes: perception and cognition.
    • In pretraining, it obtains strong item-level token perception through an itemic tokenizer and four-granularity training.
    • It uses a three-level cognition-enhanced chain-of-thought, or CoT, SFT scheme that goes from R0, which stands for perception, to R1, derivation, to R2, evolution, and finally R3, recommendation.
    • The authors use a specialize-then-unify RL strategy that first specializes by task and then unifies the models to strengthen thinking ability.
    • On the Kuaishou production multi-domain benchmark, the thinking mode consistently outperforms the non-thinking mode.
    • They find a transfer effect in which CoT supervision also improves non-thinking inference.
    • OneReason-8B and 0.8B models are planned for open sourcing.
    原文摘要PDF
  22. 360RoboticsarXiv:2606.06155

    AffordanceVLA: A Vision-Language-Action Model Empowering Action Generation through Affordance-Aware Understanding

    • It introduces a structured affordance intermediate representation to close the gap between the semantic space of VLMs and the 3D physical action space.
    • It models the task progressively through Which2Act, a visual latent prediction stage, Where2Act, a 2D affordance map stage, and How2Act, a 3D geometry stage.
    • Its MoT architecture consists of three expert modules for understanding, affordance generation, and action.
    • It addresses label sparsity with a three-stage progressive training scheme and an automatic affordance annotation pipeline.
    • On real robots, it achieves 88.3 percent on basic tasks and 82.9 percent on complex tasks, which is 38.1 percentage points better than Pi0 on complex tasks.
    • The authors hypothesize that affordances serve as a semantic anchor that preserves the visual-language capabilities of the VLM backbone.
    原文摘要PDF
  23. 359RoboticsarXiv:2606.06194

    ActiveMimic: Egocentric Video Pretraining with Active Perception

    • Camera motion in egocentric video is reinterpreted as viewpoint action rather than noise.
    • VGGT plus SAM-3D-Body reconstruct camera and wrist trajectories from a single RGB input and build a 27-dimensional unified action representation.
    • The model jointly learns active perception and manipulation with flow matching on the large-scale Ego4D dataset.
    • On four real-robot tasks, it matches or exceeds pi0 on evaluations such as Restocking at 90.1 percent and Finding at 91.7 percent.
    • The results show that active perception capability originates from egocentric pretraining rather than from robot fine-tuning.
    • Layer-level analysis confirms that camera-motion supervision promotes representation transfer from humans to robots.
    原文摘要PDF
  24. 358Computer VisionarXiv:2606.03746

    Qwen-Image-Flash: Beyond Objective Design

    • It trains a 4-NFE student model distilled from Qwen-Image-2.0 using distribution matching distillation, or DMD.
    • A counterintuitive data finding emerges: increasing diversity hurts performance, while consistent single-category data transfers better.
    • Step-wise multi-teacher guidance combines complementary strengths across teacher models and stabilizes training.
    • It achieves the best overall performance at a balanced T2I and Edit ratio, and edit supervision also improves generation quality.
    • Qwen-Image-Flash produces high-quality results with only 4 NFE in scenarios such as poster generation.
    原文摘要PDF
  25. 357Computer VisionarXiv:2606.05769

    Imagine Before You Predict: Interleaved Latent Visual Reasoning for Video Event Prediction

    • We propose an interleaved latent visual reasoning architecture that alternates between language tokens and continuous latent visual tokens.
    • Using visual-gain data curation, we built 50K high-quality training examples in which future visual hints genuinely help prediction.
    • LA-DAPO combines outcome-contrastive reward and temporal-diversity reward as a latent-awareness RL objective.
    • On FutureBench, Qwen3-VL-8B improves from 61.0 to 85.4, a 10.4-point gain over the previous best, Video-CoE.
    • On TwiFF-Bench, the average score rises from 2.44 to 3.04, and latent span usage increases adaptively as reasoning depth grows.
    • With the same data, text-only SFT reaches only 65.0, whereas interleaved latent SFT reaches 73.2, showing that the gain does not come from extra supervision.
    原文摘要PDF
  26. 356AI / GeneralarXiv:2606.05818

    Benchmarks in Leipzig

    • The benchmark contains 100 research-level mathematics problems posed by 49 mathematicians, 35 of whom participated directly in the workshop, spanning 25 subfields including algebraic geometry, combinatorics, and representation theory.
    • It uses a three-stage evaluation pipeline: Stage 1 with 5 models and 1 run each, Stage 2 with 3 models and 20 runs each, and Stage 3 with 2 models and 3 heavy-thinking runs each.
    • The number of unsolved problems drops sharply from 41 to 16 to 2, suggesting that benchmark-style practice has reached the limits of the best-performing models.
    • GPT-5.5 solves 44 problems on a single attempt, the best result, while performance varies dramatically across models, with Grok 4.3 solving only 6 problems.
    • Even repeated runs of the same model show very large performance variance, indicating a deterministic limit in mathematical reasoning.
    • It is a public benchmark based on the ScienceBench platform, and the LLM review process found and removed 3 erroneous problems.
    原文摘要PDF
  27. 355Machine LearningarXiv:2606.05254

    Flash-WAM: Modality-Aware Distillation for World Action Models

    • It identifies the structural reason why prior consistency distillation fails in video-action joint diffusion: gradient signals vanish because of an asymmetric noise schedule.
    • It selects modality-specific consistency functions, using linear-gradient scaling for actions and variance-preserving parameterization for video.
    • It delivers a 23 times speedup, reducing per-chunk latency from 8.1 seconds to 348 milliseconds and enabling real-time control.
    • It preserves an 85.5 percent success rate on RoboTwin 2.0 and 95.7 percent on LIBERO under the 1v/2a setting.
    • It achieves a 60 percent success rate on a real-world Unitree G1 humanoid, which is a substantial improvement over the naive LCM baseline at 24 percent.
    • A structural analysis of the consistency function family theoretically explains the limits of gradient scaling.
    原文摘要PDF
  28. 354AI / GeneralarXiv:2606.03883

    Reasoning Structure of Large Language Models

    • We build a scalable benchmark based on 21 two-dimensional grid logic puzzles, enabling controlled scaling studies across four difficulty levels.
    • We develop a pipeline that automatically converts unstructured text reasoning traces into atomic claim-dependency graphs.
    • We propose the reasoning efficiency metric eta, a topology-based measure that distinguishes focused reasoning from diffuse search.
    • We empirically show that token count is a poor proxy for reasoning quality, with a correlation of r = -0.05 with eta, and that extra tokens are mostly consumed by verification overhead.
    • Analysis of open-source reasoning models shows that the highest difficulty level is barely solved by any model, even with large token budgets, confirming the limits of simple compute scaling.
    • Early errors are strongly associated with inefficient traces, with r = 0.28, and eta increases when the time of error occurrence is later.
    原文摘要PDF
  29. 353Machine LearningarXiv:2606.04662

    Why Muon Outperforms Adam: A Curvature Perspective

    • A second-order Taylor expansion decomposes the loss reduction of Muon and Adam. The first-order gains are similar, but the difference appears in the second-order curvature penalty.
    • The update norms are similar, but Muon has lower NDS (normalized direction sharpness), so its curvature cost is smaller.
    • Controlled experiments with Zipf-PCFG show that data imbalance amplifies Muon's NDS advantage.
    • In the middle and later stages of training, Muon's lower NDS comes mainly from reduced within-layer curvature.
    • On heterogeneous-curvature second-order problems, Muon allocates update energy evenly across high- and low-curvature directions.
    • It is rigorously proven that when curvature heterogeneity is sufficiently large, Muon achieves a lower loss than GD.
    原文摘要PDF
  30. 352RoboticsarXiv:2606.05979

    World-Language-Action Model for Unified World Modeling, Language Reasoning, and Action Synthesis

    • It unifies WAM's future state prediction and VLA's language reasoning in a single autoregressive transformer.
    • It uses a dual structure based on meta-queries, with a World Expert for visual prediction and an Action Expert for action generation.
    • The World Expert can be disabled at inference time, which enables real-time control with 40 ms inference.
    • Test-time scaling further improves robot control accuracy.
    • It achieves state-of-the-art results on RoboTwin2.0 Clean at 92.94% and RMBench at 56.5%.
    • It also learns new tasks from cross-embodiment videos without action annotations, which points to a scalable direction for robot learning.
    原文摘要PDF
  31. 351RoboticsarXiv:2606.01955

    WALL-WM: Carving World Action Modeling at the Event Joints

    • It shifts the paradigm from chunk-centric optimization to event-grounded optimization, aligning natural boundaries across language, video, and action.
    • It builds a data ecosystem made up of event captions and cluster-balanced sampling.
    • It uses dual inference modes: Event mode for variable-length event execution and Unified mode for fixed chunks with Staircase Decoding.
    • It provides a scalable training recipe built on large-scale pretraining infrastructure with the Muon optimizer.
    • It demonstrates strong generalization across language, scene, and task settings.
    • It outperforms the existing WAM on both physical video prediction and executable control.
    原文摘要PDF
  32. 350Machine LearningarXiv:2606.06021

    OPRD: On-Policy Representation Distillation

    • It moves distillation from output-space matching to hidden-state-space matching, overcoming the structural limitations of all prior OPD variants.
    • It removes the variance problem of Monte Carlo KL estimation for large-vocabulary models such as Qwen with about 150K vocabulary by using deterministic MSE instead.
    • The teacher's layer-wise and position-wise intermediate representations, which LM-head projection would discard, are transferred directly to the student.
    • It fully closes the student-teacher gap on three competitive math benchmarks, AIME 2024, AIME 2025, and AIMO.
    • It achieves 1.44x faster training than top-k OPD on the same hardware and uses up to 54 percent less actor-update memory.
    • Its memory and distributed advantages are especially strong in expanded scenarios such as multi-model RL merging and on-policy self-distillation, or OPSD.
    原文摘要PDF
  33. 349Machine LearningarXiv:2606.06096

    OrderGrad: Optimizing Beyond the Mean with Order-Statistic Policy Gradient Estimation

    • We propose an unbiased gradient estimator for order-statistic objective functions, and by changing only the rank weights, it can express many distributional objectives such as VaR, CVaR, trimmed mean, and best-of-K.
    • It is a plug-and-play design that can be applied to existing likelihood-ratio and reparameterization updates in one line of code by reformulating them as reward transformations.
    • Top-M@K shows consistent gains over the existing Max@K, which is the special case max@K with m=1, on both pass@1 and pass@256.
    • Combining a correct-answer reward for Top-M and a length penalty for Bottom-M shortens responses far more effectively than the simple scalarization used by GRPO.
    • Choosing K, the target size, and m, the number of top ranks, controls the exploration-exploitation tradeoff, with larger K improving pass@256 while reducing pass@1.
    • It has effectively negligible overhead, with an O(N log N) sorting cost.
    原文摘要PDF
  34. 348LLMs / NLParXiv:2606.06447

    Latent Reasoning with Normalizing Flows

    • We place a normalizing flow inside the LLM backbone to provide exact likelihoods for continuous CoT while preserving left-to-right decoding and KV-cache support.
    • Continuous thought positions are handled by the NF head, while text positions are handled by the LM head, and both are processed jointly in the same causal stream.
    • The system supports both supervised likelihood training and policy-gradient optimization through a unified interface.
    • Even with Gaussian perturbations in latent space, pass@1 remains nearly unchanged, showing that the learned continuous thought space is locally smooth and functionally robust.
    • On code-generation benchmarks, it improves accuracy while reducing inference cost compared with explicit CoT and prior latent reasoning methods such as LaDiR.
    原文摘要PDF
  35. 347Computer VisionarXiv:2606.02800

    Cosmos 3: Omnimodal World Models for Physical AI

    • It is an omnimodal design based on a single mixture-of-transformers architecture that flexibly supports input and output across language, image, video, audio, and action.
    • A dual-tower layer structure and dual-stream joint attention effectively fuse multimodal information.
    • It serves both as a reasoner and a generator, and is released in Super and Nano scales.
    • It uses large synthetic datasets, such as SDG-PhyxSim, SDG-RobotSim, and SDG-DriveSim, for training.
    • According to Artificial Analysis, it was rated the best open-source text-to-image and image-to-video model, and the top policy model in RoboArena.
    • It fully releases code, checkpoints, datasets, and evaluation benchmarks under the Linux Foundation OpenMDW-1.1 license.
    原文摘要PDF
  36. 346Machine LearningarXiv:2606.06080

    On Advantage Estimates for Max@K Policy Gradients

    • The paper mathematically proves that the existing PKPO advantage estimator is unbiased for policy gradients but not centered, and addresses this with a Leave-Two-Out (L2O) baseline.
    • It derives the canonical advantage u_i - v_i for finite-batch max@K and unifies existing pass@K and max@K estimators under a single framework.
    • It is efficient with a vectorized implementation that has O(B^2) time complexity and integrates naturally into group-based RL pipelines.
    • On Llama-3.2-3B-Instruct, it reduces gradient variance by 77.4 percent relative to PKPO and improves pass@256 by 5.2 percent on Qwen2.5-Math-7B.
    • It shows consistent gains over strong post-training baselines for LLMs, including GRPO and Entropy-Adv.
    原文摘要PDF
  37. 345AI / GeneralarXiv:2606.05405

    Agents' Last Exam

    • It is a large benchmark with 55 subdomains, 13 industry clusters, and 1,490 task instances based on the O*NET and SOC 2018 occupational taxonomy.
    • Tasks are collected from projects already completed by working professionals and are evaluated automatically with deterministic scripts and structured rubrics, removing dependence on human raters.
    • The tasks are classified into three tiers by difficulty, Near-Term with 59 tasks, Full-Spectrum with 55, and Last-Exam with 36, to support agent evaluation at different levels.
    • Failure analysis shows that lack of domain knowledge, meaning understanding and approach, accounts for about 75 percent of all failures, so knowledge is the main bottleneck rather than execution ability.
    • Model choice creates about three times larger performance differences than agent harness choice, and using more resources does not always lead to better performance.
    • Even the strongest configuration, Codex plus GPT-5.5, reaches only a 2.6 percent average pass rate, which suggests that the benchmark is still far from saturated.
    原文摘要PDF
  38. 344AI / GeneralarXiv:2606.06453

    Vortex: Efficient and Programmable Sparse Attention Serving for AI Agents

    • A Python-embedded frontend allows rapid prototyping of sparse attention algorithms.
    • AI agents automatically generate and refine algorithms to achieve a 3.46x throughput improvement.
    • It accelerates GLM-4.7-Flash MLA by 4.7x and MiniMax-M2.7 229B by 1.37x.
    • It provides an efficient backend tightly integrated with modern LLM serving stacks.
    • It expands the applicability of sparse attention to emerging architectures and ultra-large models.
    原文摘要PDF
  39. 343Machine LearningarXiv:2606.06458

    In-Context Multiple Instance Learning

    • A pretrained in-context learner trained on synthetic data solves MIL tasks from only a few labels.
    • It classifies in a single forward pass at inference time and does not require gradient updates.
    • It proposes and analyzes several synthetic data generators for bag-structured data.
    • Different generators capture complementary inductive biases.
    • A mixture-generator pretrained model outperforms supervised baselines on 12 MIL benchmarks.
    原文摘要PDF
  40. 342Machine LearningarXiv:2606.06459

    Event Detection for Parameter-to-KPI Dependency Learning for AI-RAN

    • This work covers the first step of dependency learning for interference management among multiple AI control functions in AI-RAN.
    • We propose a pipeline that converts noisy continuous measurements into binary event indicators.
    • We introduce a new synthetic closed-loop traffic generator with embedded latent dependencies.
    • We evaluate the dependency recovery pipeline by formulating it as a significance detection problem.
    • We confirm that threshold calibration is a key determinant of event-detection quality.
    原文摘要PDF
  41. 341SecurityarXiv:2606.06460

    Will the Agent Recuse Itself? Measuring LLM-Agent Compliance with In-Band Access-Deny Signals

    • It proposes a Recuse Signal that asks the agent to voluntarily withdraw through the protocol's existing channel.
    • It is designed as a cooperative governance mechanism such as robots.txt, not as a security boundary.
    • It was implemented with SSH banners and PostgreSQL NOTICE messages and tested in production environments.
    • Without the signal, agents completed tasks 100% of the time, and with the signal, they withdrew 100% of the time, showing a clean contrast.
    • A hierarchical response was observed in which only the most capable model resumed entry after operator approval framing.
    原文摘要PDF
  42. 340AI / GeneralarXiv:2606.06462

    Benchmark Everything Everywhere All at Once

    • It autonomously handles the full benchmark-building pipeline, from query analysis to quality control.
    • It addresses the labor-intensive nature of existing benchmarks and the problem of rapid performance saturation.
    • It successfully generated 15 benchmarks covering text, multimodal, and domain reasoning.
    • Human evaluation and LLM-as-a-judge confirmed the high quality of the generated samples.
    • Continuous evaluation uncovered domain-specific reasoning weaknesses in current models.
    原文摘要PDF
  43. 339LLMs / NLParXiv:2606.06467

    You Only Index Once: Cross-Layer Sparse Attention with Shared Routing

    • It proposes CLSA, which shares routing indices across layers in a KV-sharing architecture.
    • A single indexer performs top-k selection once and reuses the result across all layers.
    • It amortizes routing cost while preserving the accuracy of token-sparse attention.
    • It improves all major inference bottlenecks at once, including prefilling, KV cache, and decoding.
    • It achieves a 7.6x decoding speedup and a 17.1x overall throughput gain at 128K context length.
    原文摘要PDF
  44. 338AI / GeneralarXiv:2606.06468

    Goedel-Architect: Streamlining Formal Theorem Proving with Blueprint Generation and Refinement

    • It uses a strategy that generates and refines blueprints, which are dependency graphs of definitions and lemmas.
    • Instead of recursive lemma decomposition, it avoids dead-end loops through parallel proofs and global refinement.
    • It uses DeepSeek-V4-Flash, a 284B-A13B open model, as the backbone.
    • It achieves 100 percent on MiniF2F-test, 88.8 percent on PutnamBench, 4 out of 6 on IMO 2025, and 11 out of 12 on Putnam 2025.
    • It reaches state of the art at up to 500 times lower cost than similar open-source pipelines.
    原文摘要PDF
  45. 337StatisticsarXiv:2606.06469

    How abundant are good interpolators?

    • It establishes a large-deviation principle for the generalization error when a classifier is chosen at random from the set of interpolating classifiers.
    • It shows that almost all interpolating classifiers concentrate exponentially around the same generalization performance.
    • The analysis is carried out on two natural distributions: Gaussian mixture models and logistic models.
    • In the overparameterized regime, gradient descent and linear programming outperform random interpolators.
    • The result suggests that benign overfitting in practical algorithms is a nontrivial phenomenon rather than something obvious.
    原文摘要PDF
  46. 336Machine LearningarXiv:2606.06470

    PC Layer: Polynomial Weight Preconditioning for Improving LLM Pre-Training

    • A low-order polynomial preconditioner stabilizes the singular value spectrum of the weight matrices.
    • The trained weights are merged back into the original architecture, so there is no inference overhead.
    • On Llama-1B, it improves performance over the standard transformer for both AdamW and Muon.
    • Theoretically, it guarantees geometric convergence of gradient descent when layerwise singular values are uniformly bounded.
    • It can be applied to existing LLM training pipelines without extra parameters or architectural changes.
    原文摘要PDF
  47. 335AI / GeneralarXiv:2606.06473

    MLEvolve: A Self-Evolving Framework for Automated Machine Learning Algorithm Discovery

    • It uses Progressive MCGS to enable information flow across branches, solving the isolation problem of existing tree search.
    • Retrospective Memory jointly uses domain knowledge and task experience.
    • An adaptive coding mode that separates strategy planning from code generation improves stability in long iterative loops.
    • It achieves state-of-the-art mean medal rate and valid submission rate on MLE-Bench with a 12-hour budget, which is half the standard budget.
    • It outperforms AlphaEvolve on mathematical algorithm optimization and shows strong cross-domain generalization.
    原文摘要PDF
  48. 334LLMs / NLParXiv:2606.06474

    Self-Augmenting Retrieval for Diffusion Language Models

    • It uses low-confidence tokens discarded during denoising as a lookahead signal for retrieval.
    • It is training-free, retriever-agnostic, and applicable to all discrete diffusion language models.
    • It exploits the fact that low-confidence tokens surface key entities before the output is complete.
    • It outperforms existing baselines on five multi-hop QA benchmarks.
    • It achieves up to 8x higher throughput than prior methods.
    原文摘要PDF
  49. 333Machine LearningarXiv:2606.06475

    RREDCoT: Segment-Level Reward Redistribution for Reasoning Models

    • The method addresses the delayed-reward problem in chain-of-thought reasoning through segment-level reward redistribution.
    • We use the model itself to approximate optimal reward redistribution without additional generation.
    • This greatly reduces computational overhead compared with Monte Carlo sampling.
    • We perform credit assignment by identifying CoT segments that contribute to reaching the correct answer.
    • We systematically analyze design choices for CoT segmentation and state-value estimation.
    原文摘要PDF
  50. 332Machine LearningarXiv:2606.06479

    Pretraining Recurrent Networks without Recurrence

    • It reduces RNN training to supervised learning over stepwise memory transitions and bypasses recurrent backpropagation.
    • It learns what to remember by generating memory labels with a Transformer-based encoder.
    • It captures long-range dependencies stably with an O(1) gradient path between every token pair.
    • It enables parallel training along the time axis, which has the potential to remove the bottleneck in scaling RNNs.
    • It outperforms standard BPTT on language modeling and pixel sequence modeling.
    原文摘要PDF