Beijing Jiaotong University

6 related articles

Subscribe to updates

arXiv Oct 27, 2026
Reconstruction residuals miss a drifting sensor: CLOSURE adds a second, separately learned channel-consistency score and reaches 88.13 Aff-F1 on SMD
Multivariate time series anomaly detection has long leaned on reconstruction residuals, but a residual only certifies that a window can be reproduced, not that its channels still explain one another as they do under normal operation. CLOSURE splits recoverability and cross-channel consistency into two prediction tasks trained separately on normal data: a diffusion reconstructor conditioned on a discrete memory of normal temporal bases scores the first, and a relation operator that predicts each channel only from its peers scores the second, with the two combined after calibration on normal data. Across the five public benchmarks SMD, MSL, SMAP, SWaT and PSM, its Aff-F1 exceeds the listed methods, reaching 88.13 on SMD.
Multivariate time series anomaly detection has long leaned on reconstruction residuals, but a residual only certifies that a window can be reproduced, not that its channels still explain one another as they do under normal operation. CLOSURE splits recoverability and cross-channel consistency into two prediction tasks trained separately on normal data: a diffusion reconstructor conditioned on a discrete memory of normal temporal bases scores the first, and a relation operator that predicts each channel only from its peers scores the second, with the two combined after calibration on normal data. Across the five public benchmarks SMD, MSL, SMAP, SWaT and PSM, its Aff-F1 exceeds the listed methods, reaching 88.13 on SMD.
Tianjin University Beijing Jiaotong University Shanghai Jiao Tong University Tsinghua University City University of Hong Kong
Fan Xu（First author）
CLOSURE Time-Series Anomaly Detection Relation-Closure Operator Diffusion Model
arXiv Nov 2, 2026
When students write wrong, teachers drift too: SPTA first teaches the teacher to continue from a student's half-finished reasoning
On-policy distillation trains a student on trajectories the student itself generates, but student prefixes can carry reasoning errors or patterns unfamiliar to the teacher, and the teacher's token-level supervision becomes less reliable as prefixes grow longer. Before distillation begins, SPTA trains the teacher on student prefixes of random lengths, rewarding correct final answers, so the teacher learns to recover from student-induced states rather than imitate the student's reasoning; standard on-policy distillation then proceeds with the distillation objective unchanged. On mathematical reasoning benchmarks, students trained with SPTA average better than under standard on-policy distillation and better than under on-policy distillation with task-level RLVR-adapted teachers.
On-policy distillation trains a student on trajectories the student itself generates, but student prefixes can carry reasoning errors or patterns unfamiliar to the teacher, and the teacher's token-level supervision becomes less reliable as prefixes grow longer. Before distillation begins, SPTA trains the teacher on student prefixes of random lengths, rewarding correct final answers, so the teacher learns to recover from student-induced states rather than imitate the student's reasoning; standard on-policy distillation then proceeds with the distillation objective unchanged. On mathematical reasoning benchmarks, students trained with SPTA average better than under standard on-policy distillation and better than under on-policy distillation with task-level RLVR-adapted teachers.
Tencent (China)Beijing Jiaotong University
Xue Zhang（First author）
Student-Prefix Teacher Adaptation On-Policy Distillation Reinforcement Learning with Verifiable Rewards Mathematical Reasoning
arXiv Oct 12, 2026
A chest X-ray agent stops calling tools by default: it first asks whether evidence is needed, then blocks unusable measurements
The chest X-ray agent X-Ragent splits "should I seek evidence" from "is this measurement usable": the policy invokes region zoom, anatomical parsing and knowledge lookup only when needed, while quantitative measurements such as the cardiothoracic ratio enter reasoning only through an independent validity gate. Training runs in two stages, agentic supervised fine-tuning on 7,715 structured diagnostic trajectories followed by GRPO with an accuracy-gated reward that applies process terms only to correct trajectories. On the fully held-out ChestAgentBench, accuracy rises from 63.1% for the strongest baseline to 71.7%, with gains across all seven interpretation capabilities.
The chest X-ray agent X-Ragent splits "should I seek evidence" from "is this measurement usable": the policy invokes region zoom, anatomical parsing and knowledge lookup only when needed, while quantitative measurements such as the cardiothoracic ratio enter reasoning only through an independent validity gate. Training runs in two stages, agentic supervised fine-tuning on 7,715 structured diagnostic trajectories followed by GRPO with an accuracy-gated reward that applies process terms only to correct trajectories. On the fully held-out ChestAgentBench, accuracy rises from 63.1% for the strongest baseline to 71.7%, with gains across all seven interpretation capabilities.
Alibaba Group (China)National University of Singapore Beijing Jiaotong University City University of Hong Kong Santa Clara University Carnegie Mellon University
Yi Wang（First author）
X-Ragent Tool-Augmented Reinforcement Learning Rad-CoT GRPO Diagnostic Reasoning
arXiv Oct 14, 2026
Music generation stops chasing generic appeal: distilling a listening history into an aesthetic profile makes the same track fit the listener
PAG-Agent distills a listener's favorite songs and contextual profile into an explicit Personal Aesthetic Profile that a planning model turns into generation instructions, raising target-user consistency by roughly 9 to 19 percent over direct conditioning on Personal-100.
Beijing Jiaotong University Beijing University of Technology
Yifeng Mao（First author）
PAG-Agent Aesthetic Profile Distillation PA-Eval Personal-100 Music Generation
arXiv Oct 12, 2026
Autonomous driving planners stop learning from one score: reflective guidance folded into a single pass, training time cut by over 90%
Autonomous-driving vision-language-action models have learned from a scalar trajectory score that flags a bad plan but cannot say whether scene understanding, behavior selection, or trajectory execution went wrong. Reflect-VLA treats reflective guidance, which pairs evaluation diagnoses with corrective suggestions, as training-time privileged information and internalizes it through dual-path on-policy self-distillation: risky trajectories align with the teacher's token-level corrective distribution, while reliable ones are preserved by self-imitation. It reaches competitive planning performance on NAVSIM v1/v2 and Bench2Drive, cuts training time by over 90% versus reinforcement learning, and generates trajectories in a single pass at inference without guidance or iterative refinement.
Autonomous-driving vision-language-action models have learned from a scalar trajectory score that flags a bad plan but cannot say whether scene understanding, behavior selection, or trajectory execution went wrong. Reflect-VLA treats reflective guidance, which pairs evaluation diagnoses with corrective suggestions, as training-time privileged information and internalizes it through dual-path on-policy self-distillation: risky trajectories align with the teacher's token-level corrective distribution, while reliable ones are preserved by self-imitation. It reaches competitive planning performance on NAVSIM v1/v2 and Bench2Drive, cuts training time by over 90% versus reinforcement learning, and generates trajectories in a single pass at inference without guidance or iterative refinement.
University of Macau Beijing Jiaotong University Nanyang Technological University Tsinghua University
Yuechen Luo（First author）
Reflect-VLA Dual-Path On-Policy Self-Distillation Reflective Guidance Trajectory Prediction End-to-End Autonomous Driving
arXiv Oct 2, 2026
Multimodal Flow unifies language and vision as continuous hyperchunks under one flow-matching backbone, with MF-1 scoring 0.821 on GenEval and 83.44 on DPG-Bench at 150B tokens
The work introduces Multimodal Flow, which encodes text blocks and images as ordered continuous hyperchunks and models both modalities in their own embedding spaces with a shared chunk-causal Flow Matching backbone, instantiated as MF-1: continued pretraining improves multimodal modeling across 0.6B, 1.2B, and 1.6B scales, and the 1.6B model reaches 0.821 on GenEval and 83.44 on DPG-Bench with only 150B pretraining tokens and an average of 75.3 across VQAv2, MMBench, and POPE, outperforming representative hybrid and discrete models under matched data, optimization, and parameter budgets.
The work introduces Multimodal Flow, which encodes text blocks and images as ordered continuous hyperchunks and models both modalities in their own embedding spaces with a shared chunk-causal Flow Matching backbone, instantiated as MF-1: continued pretraining improves multimodal modeling across 0.6B, 1.2B, and 1.6B scales, and the 1.6B model reaches 0.821 on GenEval and 83.44 on DPG-Bench with only 150B pretraining tokens and an average of 75.3 across VQAv2, MMBench, and POPE, outperforming representative hybrid and discrete models under matched data, optimization, and parameter budgets.
Huazhong University of Science and Technology Beijing Jiaotong University Horizon Robotics (China)
Hongyuan Tao（First author）
Multimodal Flow MF-1 Flow Matching Text-to-Image Generation Visual Question Answering

You’re all caught up

Reconstruction residuals miss a drifting sensor: CLOSURE adds a second, separately learned channel-consistency score and reaches 88.13 Aff-F1 on SMD

When students write wrong, teachers drift too: SPTA first teaches the teacher to continue from a student's half-finished reasoning

A chest X-ray agent stops calling tools by default: it first asks whether evidence is needed, then blocks unusable measurements

Music generation stops chasing generic appeal: distilling a listening history into an aesthetic profile makes the same track fit the listener

Autonomous driving planners stop learning from one score: reflective guidance folded into a single pass, training time cut by over 90%

Multimodal Flow unifies language and vision as continuous hyperchunks under one flow-matching backbone, with MF-1 scoring 0.821 on GenEval and 83.44 on DPG-Bench at 150B tokens