Machine Learning
ChemOPD: Multi-Teacher On-Policy Distillation for Multi-Task Chemical Reasoning
Yaoyao Xu, Xinjian Zhao, Xiaozhuang Song et al. (+2 more)
Large
language models are increasingly expected to support diverse chemical
reasoning capabilities within a unified model. One approach is to develop specialized capabilities separately and consolidate them through multi-teacher on-policy
distillation, but this raises two questions: how should specialization be organized, and how should specialist guidance be integrated? We introduce ChemOPD, which addresses both. We estimate task affinities from supervised fine-tuning gradients and solve a constrained mixed-integer program(MIP) to construct partially overlapping specialist groups. During
distillation, we retain a generalist teacher trained on all tasks so that specialist guidance supplements rather than replaces its supervision. Our anchor-residual objective gradually increases the routed specialist's contribution on student-generated responses. On ChemCoTBench, affinity-guided specialization produces task-dependent gains over the generalist teacher and improves several capabilities beyond semantic task grouping. Yet stronger teacher-side performance does not automatically yield stronger students: with the same specialists and routes, anchor-residual OPD improves most reported metrics over specialist-only
distillation and realizes a larger share of the available teacher gains. These results highlight specialization and capability integration as connected but distinct design problems in chemical
reasoning.
Read the full abstract ▾
All authors Yaoyao Xu, Xinjian Zhao, Xiaozhuang Song, Xuemin Chen, Tianshu Yu
Machine Learning · Cross-listed from stat.ML
An Active-Bottleneck Mechanism for Weak-to-Strong Generalization
Mohammad Zeinalpour and Amir Najafi
Weak-to-strong generalization (W2SG) occurs when a student trained on a teacher's predictions outperforms that teacher. We study when this happens under fully converged, ridgeless two-stage learning, with no early stopping, no explicit regularization, and no assumption that the student is more expressive than the teacher. In two-stage linear regression, a teacher is fit from $n$ labeled examples and a student is trained solely on the teacher's predictions on $m$ fresh, unlabeled inputs. Although both stages share the same hypothesis class and the same training rule, we show that the student outperforms the teacher exactly when $m$ lies in an explicit intermediate range: too few pseudo-labels leave the student without enough signal, too many let it inherit the teacher's noise. Under power-law covariance, we derive this range in closed form as a function of the spectral decay and noise level, including regimes where the improving region splits into two disjoint intervals of $m$. We then study a random-feature model in which the student has strictly more features than the teacher, and identify two regimes, again given by explicit thresholds: one where improvement occurs only for $m$ in a bounded interval, and one where it occurs only once the student width $N_S$ exceeds an explicit threshold. Both regimes are governed by a single "active-bottleneck" principle: whichever of $m$ or $N_S$ is scarcer controls how much teacher error is filtered out, while increasing the other resource only reduces estimation noise. Together, these results show that finite data and finite width can themselves regularize a two-stage learner, with no explicit mechanism doing so.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.CV
CLIMB-flow: Coupled Linear Inverse posterior sampling via Multiscale-Based flow
Zeqiu Yu, Ruizhi Yuan and Mathews Jacob
Diffusion models are now widely used in
Bayesian inverse problems in imaging as priors, where latent
diffusion models are often used for larger scale problems to keep the computational complexity and model-size manageable. Unfortunately, the auto-encoder based compression results in loss of spatial detail. In addition, the optimization is converted to a non-linear problem. In this paper, we introduce a posterior sampling
algorithm customized for the pyramidal/cascaded architecture, which relies on a coarse to fine hierarchical strategy to generate images in the pixel domain. We present CLIMB-Flow which alternates between three steps: an end-point estimation from the current coarse and noisy image, data-consistent update of the clean image, and re-noising it back to the level the network expects. Together these steps sample the posterior at that scale using an approximate Gibbs sampling from two conditional distributions. Experiments on ImageNet, CelebA, AFHQ and fastMRI span inpainting, deblurring, super-resolution and accelerated MRI, with PSNR gains of 1.37-7.66 dB over the strongest competing method on CelebA and pixel-domain reconstruction up to 512x512.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.CV
One Attack to Fool Them All: Highly Transferable Black-Box Adversarial Attacks on Frontier MLLMs
Sen Nie, Jie Zhang, Zhongqi Wang et al. (+2 more)
Adversarial attacks have long posed a fundamental threat to machine learning systems. As
multimodal large
language models (MLLMs) rapidly evolve and become widely deployed, assessing their vulnerability to such attacks is essential for their safe use. In this work, we investigate whether a single
adversarial image can consistently mislead diverse frontier MLLMs in black-box settings. We propose O-Attack, a highly transferable black-box attack framework. This framework builds on our insight that surrogate models contain a broad, high-level, cross-modally aligned semantic space. This space extends beyond final-layer outputs and provides multiple semantically consistent representations that remain underexploited by existing attacks. Within this space, O-Attack anchors aligned representations, progressively broadens semantic conditions, and optimizes perturbations through semantic consensus to promote consistent target
alignment. By fully exploiting this space with the same surrogate models as M-Attack, O-Attack raises attack success rates on GPT-5.4 (29.1% to 77.2%), Claude-4.6 (42.8% to 81.6%), and Gemini-3.1 (38.2% to 80.9%). Extensive experiments across 24 MLLMs show that O-Attack outperforms six state-of-the-art methods in black-box transferability, with consistent effectiveness across
prompts and improved efficiency and imperceptibility. This work exposes the practical safety risks posed by black-box
adversarial attacks against frontier MLLMs, underscoring the need for more rigorous robustness evaluation and more effective defenses.
Read the full abstract ▾
All authors Sen Nie, Jie Zhang, Zhongqi Wang, Shiguang Shan, Xilin Chen
Machine Learning · Cross-listed from cs.RO
Achieve What You Imagined: Learning to Align Actions with Visual Plans
Yuheng Qiao, Ziran Wei, Xiaohan Wang et al. (+5 more)
World-action models can jointly predict future visual observations and
robot actions. However, discrepancies may exist between their visual predictions and the consequences implied by generated actions. We observe that WAMs can often generate visually plausible task-completion outcomes before producing action sequences that reliably achieve them. Consequently, we treat the WAM-generated visual prediction as a goal-conditioned visual proposal rather than a directly executable plan. We use a frozen action-conditioned world model to predict action-conditioned consequences and construct feedback based on consistency between the two future predictions and
alignment with the terminal goal. Leveraging this feedback, we employ Flow Policy Optimization (FPO) to optimize the action head of the WAM. This framework avoids online
robot interaction and additional training of task-specific reward models. Across four real-world UR5 manipulation tasks, our method increases the mean success rate from 43.4% to 75.1%, compared with 61.4% for $\pi_{0.5}$. These results show that cross-model prediction discrepancy can provide useful feedback for improving
robot policies under the evaluated manipulation tasks. Website: https://imagine-to-achieve.github.io/
Read the full abstract ▾
All authors Yuheng Qiao, Ziran Wei, Xiaohan Wang, Daqiang Guo, Yichen Luo, Zhibo Pang, Peng Zhou, Sichao Liu
Machine Learning
dOPT: Differentiating Conic Optimization via Geometric Reduction
Fengyu Yang, Connor W. Magoon, Tyler Watts et al. (+1 more)
Optimization layers enable the incorporation of structured constraints and decision problems into learning systems. Training such systems requires differentiating through the embedded optimization problem, which can be challenging for general conic programs. We introduce dOPT, a solver-agnostic framework that, rather than differentiating the full conic formulation, reduces it at a computed primal-dual solution to an equality-constrained quadratic program that preserves the reference solution and its first-order sensitivity. The reduction captures the local first- and second-order conic geometry relevant to differentiation and remains well defined at singular configurations. Computing solution derivatives then requires a single symmetric linear solve, independently of the forward solver. We derive explicit reductions for convex NLPs, QPs, SOCPs, and SDPs. Numerical experiments validate the computed gradients and show favorable backward-pass scalability, with substantial speedups over existing differentiable conic optimization methods as problem size increases.
Read the full abstract ▾
All authors Fengyu Yang, Connor W. Magoon, Tyler Watts, Shahar Z. Kovalsky
Machine Learning · Cross-listed from math.OC
Annealed Sinkhorn with Momentum: Certified Unregularized Optimal Transport in Linear Memory
Samuel J. K. Chin and Maximilian Schiffer
We characterize Bregman Douglas-Rachford splitting (BDRS) for unregularized discrete optimal transport and develop an anytime primal-dual certificate in linear
memory. We first establish that BDRS coincides with warm-started Inexact Proximal point method for exact Optimal Transport (IPOT) using a single inner Sinkhorn iteration. By eliminating the primal transport plan from the updates, we derive an equivalent dual formulation that reveals BDRS as annealed Sinkhorn under an implicit inverse-linear temperature schedule, with an additional log-scaling momentum term and a cooler kernel. While this explains the role of the temperature parameter in BDRS as an initial temperature, it also reduces the solver's
memory requirement from quadratic to linear. Utilizing this annealing perspective, we introduce overrelaxed BDRS, which combines annealing and overrelaxed scaling within a single recursion. We derive a primal-dual certificate for both methods that can be evaluated in linear
memory without transport plan construction, thus providing a computable stopping rule. On the DOTmark
benchmark, combining momentum with the cooler kernel produces substantially smaller optimality gaps than annealed Sinkhorn under the same schedule. For pixel-level color transfer between $1024\times1024$ images, BDRS attains a lower repaired transport cost than MDOT-TNT with a 9$\times$ speed up, reaching a relative duality gap of $1.59\%$ in 24 minutes. We further demonstrate a color transfer with $4238\times2365$ images, yielding 10 million pixels per image and approximately one hundred trillion implicit transport entries, reaching a best relative duality gap of $2.41\%$ and $2.80\%$ within 35 hours in each direction on a single NVIDIA L40S GPU.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.SE
Identical Runs, Different Results: Benchmarking AI Coding Agents on Open-Weight Models
Eduardo Ariño de la Rubia and Szilard Pafka
Repeated runs of the same coding
agent are known to give different
benchmark scores. We ask what that variation means for a team running an
agent on its own task, by intensive replication on one machine-learning task: an
agent improves the training code of an XGBoost classifier for airline delays, and a holdout it never sees scores the result. Across 584 runs, we compare six
agents on six open-weight model endpoints, run six
agent-model pairings 52 times each under fixed settings, and repeat three of them on a larger model from the same family. Identical runs of one pairing varied more than the pairings differed from one another, so comparisons of a few runs ranked them unreliably; resolving the
agent differences we observed would take tens to more than a hundred runs of each. Runs on the larger model scored clearly higher, but by less than one run-to-run standard deviation, and the gap was more than twice as large with one
agent as with the others. Fewer than one run in twenty broke the task's data rules, but those runs held the highest scores. Rejecting those runs first and keeping the best compliant result among a few attempts reliably improved the delivered model, even though a few runs could not rank the
agents. On flights from a later year, the delivered models kept only a third of their gain over the starting code. At list prices, cost differed more than twentyfold between two
agents on the same model, mostly through the
prompt cache.
Agents and models should be evaluated as pairings, over repeated attempts, with compliance reported beside quality. Data, code and every delivered program: https://github.com/earino/identical-runs-different-results
Read the full abstract ▾
Machine Learning · Cross-listed from cs.SD
Controlling Speaking Rate in Autoregressive TTS via Activation Steering
Francesco Verdini, Antonis Asonitis, Aref Farhadipour et al. (+4 more)
Autoregressive text-to-
speech (TTS) systems synthesize natural
speech but, once trained, offer little control over speaking rate. We show that speaking rate can be steered at inference time, without retraining, by clamping a single decoder block's activation along a discovered speed axis. A decoder-block analysis recovers the rate axis, a neutral operating point, and a per-step intensity scale; at inference, the activation's projection onto this axis is set to a fixed scalar. Learning this direction from synthetically time-stretched and time-compressed
speech yields rate control that largely preserves speaker identity, generalizes across model architectures, and maintains high naturalness in objective and human evaluations. Unlike standard additive steering, which breaks at the slow extreme, clamping remains stable on all three systems tested; at moderate targets, the better rule depends on the model. Finally, we show that rate information is decodable across layers but causally steerable only within a mid-depth window, and demonstrate the effectiveness of our approach on the public Seed-TTS-Eval
benchmark.
Read the full abstract ▾
All authors Francesco Verdini, Antonis Asonitis, Aref Farhadipour, Marzieh Razavi, Pierre-Edouard Honnet, Vijeta Avijeet, Juan Pablo Zuluaga Gomez
Machine Learning
MinkowskiPE: Minkowski Positional Encoding for Spatiotemporal Perception
Yuhao Li, Louie Hong Yao, Tianyi Shi et al. (+4 more)
Modeling spatiotemporal coupling is a key challenge in building physical intelligence across scales, from microscopic to macroscopic. Existing models capture such structure broadly through physics-motivated dynamical formulations or learning-motivated architectures. The former provide stronger priors but may constrain flexibility, whereas the latter are more flexible but leave the spatiotemporal coupling largely implicit. We therefore seek an approach that combines flexible learning with an explicit
geometric bias for jointly modeling time and space. To this end, we propose Minkowski Positional Encoding (MinkowskiPE), which uses joint temporal and spatial coordinates to parameterize Lorentz transformations applied to query and key features. With MinkowskiPE, the query-key
attention score depends on position only through the relative spacetime displacement between the two
tokens and is therefore invariant to global translation of the coordinates. This paradigm retains the standard dot-product
attention interface and remains compatible with efficient
attention implementations. We evaluate MinkowskiPE on microscopic molecular dynamics and macroscopic
video prediction tasks, achieving the best results on all nine multi-trajectory molecular evaluations and reducing KTH
video-prediction MSE by 9.9% relative to the best baseline while using roughly one-tenth as many parameters.
Read the full abstract ▾
All authors Yuhao Li, Louie Hong Yao, Tianyi Shi, Hanqun Cao, Hongxia Hao, Zhen Zhao, Shengchao Liu
Machine Learning
Xiangyang Wang, Bingxiang He, Zeyuan Liu et al. (+13 more)
Reward models underpin the
alignment of large
language models, yet the dominant designs reduce each
prompt–response pair to a point estimate or to a distribution from a fixed parametric family. This is at odds with human preference, which is inherently
multimodal: the same response can be reasonably judged in many ways, and no single family covers all of them. To better fit this structure, we introduce DRM, a
Diffusion Reward Model that recasts reward modeling as conditional density estimation over $p(\mathbf{r}\mid x,y)$. Conditioned on a frozen LLM encoder, a lightweight
Diffusion Transformer denoises Gaussian noise into a reward vector, placing no parametric assumption on the output distribution and naturally representing its
multimodal structure. A single architecture handles both multi-attribute regression and pairwise preference data, and at inference $N$ samples form an empirical reward distribution that can be aggregated into a scalar, a variance, or
quantiles. Across five
benchmarks, DRM matches or surpasses baselines under matched data and backbone, stays competitive with much larger discriminative, distributional, and generative RMs despite its modest training scale, and recovers
multimodal reward structure where conventional heads collapse to a point. Uncertainty-aware rejection and lower-confidence-bound (LCB) aggregation further demonstrate that DRM can exploit distributional information beyond a scalar reward to improve reward-model decisions. Downstream RLHF experiments additionally show that using DRM as the training-time reward leads to improved policy performance, directly validating the practical benefit of
diffusion-based reward modeling for RLHF training.
Read the full abstract ▾
All authors Xiangyang Wang, Bingxiang He, Zeyuan Liu, Jiaze Wang, Ziqing Qiao, Yuxin Zuo, Huan-ang Gao, Cheng Qian, Wenbin Zhang, Ran Li, Youbang Sun, Ning Ding, Yuanchun Shi, Zhiyuan Liu, Chaojun Xiao, Chun Yu
Machine Learning · Cross-listed from quant-ph
Autonomous phase discovery
Shiyu Zhou, Yuxuan Zhang, Sebastian Wetzel et al. (+2 more)
Understanding
quantum phases of matter has long relied on physicists' intuition and mathematical tools such as
symmetry and
topology. Remarkably successful as these approaches have been, they provide no universal way to explore a
Hamiltonian space whose organizing principle is not known in advance. In this work, we introduce a fully autonomous system combining differentiable programming and unsupervised learning for
quantum phase discovery. The search evaluates ground-state data along an adaptive trajectory rather than on a predetermined parameter grid. We demonstrate the system with three different solvers and
benchmark it against random sampling at equal ground-state-evaluation budgets. On a generalized cluster chain hosting up to $200$ distinct phases, the search finds up to $25$ more phases at the same budget, and matches random sampling given thirty times its budget. On a $50$-parameter Chern
insulator, it reaches sectors not obtained by the simple harmonic constructions considered here, in a family whose inverse problem remains open, while recovering all sectors found by sampling. Our results establish autonomous, gradient-driven exploration of
Hamiltonian space as a practical route to discovering
quantum phases without phase labels or a prescribed target phase.
Read the full abstract ▾
All authors Shiyu Zhou, Yuxuan Zhang, Sebastian Wetzel, Roger Melko, Xiu-Zhe Luo
Machine Learning
Binding Multiple Modalities via Multimodal Wasserstein Barycenter
Xiaole Tang, Jiayi Xu, Xiang Gu et al. (+2 more)
Multimodal learning beyond two modalities commonly leverages a specific modality (e.g., text) to bind other modalities. However, how to establish a more balanced representation space that approximates shared semantics while respecting the holistic geometry of $n$-modal data remains challenging. In this work, we present BaryBind, which aims to transport the specific modality towards the Wasserstein barycenter (WB) optimized across all modalities and introduces a volumetric
alignment objective to establish a unified semantic space around the WB
embedding. Specifically, we project specific modalities to the WB, which minimizes the average Wasserstein distances to
multimodal distributions and serves as the anchor for subsequent
alignment. We then construct a barycenter simplex, whose volume is taken as a similarity metric for global
alignment centered at the WB. Experiments show that BaryBind achieves competitive performance in text-
video-audio
retrieval, classification, videoQA, and cross-modal generation tasks, along with robustness under modality absence and scalability to more than three modalities. Code is released at https://github.com/xl-tang3/BaryBind.
Read the full abstract ▾
All authors Xiaole Tang, Jiayi Xu, Xiang Gu, Yan Yang, Jian Sun
Machine Learning · Cross-listed from cs.SE
RAISE: Reinforcing Access Control Policy Synthesis in LLMs via Symbolic Evaluation
Yingming Zhou, Adarsh Vatsa and William Eiers
Translating natural-
language access-control requirements into policies requires careful
reasoning about permissions, constraints, and exceptions, and even frontier LLMs often produce policies that violate the intended authorization semantics. We construct CedarInstruct, to our knowledge the first
dataset that supports both training and semantic evaluation for formally verifiable Cedar policy synthesis. It contains 5,800 scenarios across 44 domains and 1,408 representing a single synthetic organization, each with a verified target policy and an executable verification plan. On this data we introduce RAISE, which trains policy synthesizers from formal verification in two stages, verified supervised fine-tuning (SFT) followed by a reinforcement learning (RL) stage that learns from verifier signal. We find that SFT succeeds largely by letting models express authorization logic they already have, since untrained models rarely write valid Cedar but often reason correctly when they do. After SFT, how the verifier's information is used matters more than how much of it is used. Of six RL instantiations that consume progressively richer verifier signal, only RAISE-OC improves meaningfully on SFT; it turns failed checks and symbolic counterexamples into guided exploration and learns from the result with off-context GRPO. With about 5.4K verified scenarios and LoRA fine-tuning, RAISE-OC trains Qwen3.5-9B to surpass zero-shot GPT-6 Astra and Claude Opus 5 by 13.33 and 16.26 percentage points in semantic success on held-out scenarios, and training transfers to the independently constructed CedarBench.
Read the full abstract ▾
Machine Learning
Do We Really Need KL Divergence for On-Policy Distillation of Large Language Models?
Wenze Lin, Jiyuan Long, Jiale Zhao et al. (+11 more)
Since the advent of knowledge
distillation, KL divergence has been the standard loss in
distillation. Recently, on-policy
distillation (OPD) has emerged as an efficient post-training paradigm for LLMs. As a
distillation method, OPD naturally inherits KL divergence as its standard loss. However, in this work, we find that KL divergence may not be necessary for OPD. We show that simply preserving the update direction is sufficient for effective OPD. As long as the update direction is toward the teacher, OPD works. More precisely, it is not the direction of every
token, but the direction of a small subset of
tokens where the teacher and student disagree strongly. We first show that simply assigning a reward of (+1) to
tokens where the teacher probability is higher than the student probability and (-1) where it is lower, which merely encourages updates toward the teacher, reproduces almost the same training mode as OPD with reverse KL. We further show that only the direction of a small subset of
tokens with large teacher-student disagreement is critical, and training works as long as their update direction is toward the teacher, even if other
tokens are pulled away from the teacher. And as an application of these findings, we introduce Consensus Multi-Teacher On-Policy
Distillation (C-MOPD) to improve Multi-Teacher On-Policy
Distillation (MOPD). Unlike MOPD, which routes each sample to a single teacher and may cause capability conflicts across domains, C-MOPD lets every sample be supervised by all teachers. Experiments show that C-MOPD consistently outperforms MOPD on both math and code
benchmarks. Our code is available at https://github.com/LeapLabTHU/KL-Free-OPD.
Read the full abstract ▾
All authors Wenze Lin, Jiyuan Long, Jiale Zhao, Shenzhi Wang, Xitai Jiang, Ce Luo, Rui Lan, Qianli Ma, Fukang Wen, Hui Wu, Liyuan Chen, Shuoling Liu, Jiangpeng Yan, Gao Huang
Machine Learning
Surprising Success, Repeated Failure: Entropy-Guided Credit Assignment for Exploration in LLM Reasoning
Woongyeong Yeo, Minki Kang, Chanuk Lee et al. (+3 more)
Reinforcement learning with verifiable rewards (RLVR) enhances
reasoning in large
language models (LLMs) through outcome-level feedback, yet recent approaches to finer-grained credit assignment often require auxiliary models, additional sampling, or privileged information. Although policy
entropy provides a readily available signal, prioritizing uncertain positions under both reinforcement and penalization concentrates penalties where failed responses still retain alternatives for recovery, which can suppress opportunities for exploration. To address this, we introduce Entropic Advantage Policy Optimization (EAPO), an
entropy-guided credit assignment method that treats success and failure asymmetrically. Specifically, motivated by the observation that success under uncertainty is less repeatable while confident failures tend to recur, EAPO couples normalized policy
entropy with the sign of the response advantage to reinforce surprising success and correct repeated failure. It assigns stronger reinforcement to high-
entropy decisions in successful responses and stronger penalties to low-
entropy decisions in failed responses, while attenuating penalties at uncertain positions to preserve opportunities for recovery. By redistributing the response advantage across
tokens, EAPO derives
token-level credit directly from existing rollout signals without additional supervision. We validate EAPO on a range of
reasoning tasks across both base and
reasoning backbones, demonstrating that it achieves the best overall performance. We further show that EAPO promotes more effective exploration, broadening problem coverage and generating more diverse candidate answers.
Read the full abstract ▾
All authors Woongyeong Yeo, Minki Kang, Chanuk Lee, Sangwoo Park, Jinheon Baek, Sung Ju Hwang
Machine Learning
Selecting Diverse SFT Traces Improves Post-RL Generalization
Dylan Zhang, Mingyuan Wu and Jinning Li
Verified solutions are not equally useful for preparing
reasoning models for reinforcement learning (RL). We present a comprehensive study of route diversity, the variation in the sequences of
reasoning steps in supervised fine-tuning (SFT) data, and propose a lightweight, rule-based fingerprint to select for it. From one pool at one budget, with matched training recipes and checkpoints, selecting diverse rather than similar routes improves post-RL problem coverage across puzzles and mathematics, including on problems harder than those seen in either training stage. In synthetic experiments, route-diverse SFT improves OLMo3-7B's pass@8 by 16.9 points on environments held out from SFT. In a single-model condition, where one model writes every candidate, diverse selection gains up to 6.2 points of mean pass@8 across 10 mathematics
benchmarks. Pre-RL diagnostics suggest why: diverse SFT can produce both successful and failed attempts on more
prompts despite slightly lower mean accuracy, giving group-relative RL more
prompts with a learning signal. On 3 open-source corpora, our CPU-only selector, without model calls, outperforms more expensive alternatives in every comparison of mean post-RL performance. These results identify
reasoning-route diversity as a practical criterion for selecting SFT data that better prepares models for RL.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.SD
Tokens Change, Structure Endures: Spectral Watermarking for Generated Speech
Kanghwi Lee, Kyeongseok Jeong and Jeongmin Liu
Watermarking is a promising tool for establishing the provenance of AI-generated
speech. While many neural audio watermarking methods rely on a separately trained watermark generator,
token-level watermarking is a training-free alternative that operates directly during generation. Its main weakness is retokenization: decoding generated
speech to a waveform and encoding it again can change
token identities and erode the watermark. To make the watermark robust to these changes, we propose Redwing, REtokenization-Durable Watermarking IN Generation. It builds a graph from the
token substitutions observed under retokenization, whose
Laplacian yields a basis that assigns similar values to
tokens likely to substitute for one another. Over this basis,
embedding and detection functions are jointly optimized to preserve watermark signal through retokenization while limiting
embedding distortion and detector variability on unwatermarked
speech. On the Moshi full-duplex system, after eight consecutive passes of Mimi resynthesis, Redwing achieves 80.7% TPR at a calibrated 1% FPR, compared with 8.3% for KGW and at most 7.3% for WMAR. It also has the highest TPR after eight passes through three other neural codecs (77.5-93.0%), and the gains generalize to TTS models at a
speech-quality cost close to that of KGW. These results show that retokenization is not merely a source of noise: its transition structure can be exploited as a design principle for robust
token-level watermarking.
Read the full abstract ▾
Machine Learning
DEALS: Decentralized Expertise-Aware Load Serving for Multi-Agent LLM Systems
Jingjuan Huang, Wenbin Wang, Yanchuan Yin et al. (+2 more)
Multi-
agent systems (MAS) have recently emerged as an effective approach for coordinating large
language model (LLM)-based
agents to solve complex tasks through structured interactions. In practice, MASs often handle a stream of heterogeneous and complex tasks, requiring
agents to decompose each task and then self-organize and self-evolve to adapt to incoming tasks while sharing execution resources. However, most early approaches to MASs rely on centralized controllers or fixed coordination patterns, which can limit scalability or adaptability. In contrast, existing decentralized and dynamic MASs often require training dedicated routers or invoking LLMs for
agent selection, resulting in substantial computational costs and coordination overhead. To address these challenges and enable efficient task-level self-organization and self-
evolution for task- and workload-level collaboration, we propose Decentralized Expertise-Aware Load Serving (DEALS), a decentralized and low-complexity framework that enables
agents to self-organize and dynamically route concurrent tasks for processing. Specifically, each
agent maintains local queues of incoming tasks, and its router decides whether to process a task locally or forward it to a neighbor based on differences in backlog and success rate. Meanwhile, executors process independent tasks concurrently within and across
agents, and partially solved tasks can be resumed by other
agents. Experiments show that DEALS not only improves performance along multiple dimensions (e.g., answer accuracy and task throughput) in both homogeneous and heterogeneous
agent pools, but also balances
agent expertise and workload in a self-organized manner, enabling effective decentralized coordination.
Read the full abstract ▾
All authors Jingjuan Huang, Wenbin Wang, Yanchuan Yin, Alvaro Velasquez, Jia Liu
Machine Learning · Cross-listed from q-fin.PM
Taming the Greeks: Option Portfolios with Inductive Biases
Wee Ling Tan, Stephen Roberts and Stefan Zohren
We present an end-to-end deep learning framework for systematic options trading that directly embeds hedging behavior through explicit control of portfolio-level risk exposures. While neural networks trained to optimize risk-adjusted performance have been shown to outperform traditional rules-based strategies, such approaches remain agnostic to the sensitivities of the resulting portfolios with respect to specific underlying risk factors. We propose a general training objective that combines a performance-driven loss with a differentiable risk-sensitivity penalty, enforcing neutrality to selected risk dimensions. Unlike reinforcement learning methods that approximate optimal hedging policies via simulated
market dynamics, our framework operates entirely on historical data and jointly optimizes risk-adjusted returns and targeted risk constraints in a single learning problem. We instantiate the framework on static delta-neutral straddle portfolios with the penalty directed at first-order directional exposure, and evaluate two penalty variants – an exposure-normalized penalty and a Greek-ratio drift penalty. Empirical results on Nasdaq 100 equity options demonstrate that appropriately calibrated regularization simultaneously improves out-of-sample risk-adjusted performance relative to an unregularized baseline while reducing realized directional exposure.
Read the full abstract ▾
Machine Learning
Beyond Fixed Features: Architecture-Dependent Sensitivity to Node Representations under Heterophily
Priyanath Maji, Sidharth Gaur and Rajavinoth Paul Durai
Graph Neural Networks (GNNs) perform well on homophilic graphs but struggle in heterophilic settings, where connected nodes often carry dissimilar labels. Existing evaluations typically compare architectures under a fixed node-feature representation, leaving unclear whether conclusions about heterophily robustness remain stable as the input representation changes. We address this question by constructing parallel feature variants of two large-scale heterophilic
benchmarks, Roman-Empire and Amazon-Ratings, pairing each graph with representations ranging from static fastText vectors to contextual
Transformer embeddings and evaluating seven GNN architectures across these representations. We find that the effect of representation varies across architectures: on Roman-Empire, the contextual gain ranges from 2.38 percentage points for GCN-sep to 13.67 points for GAT, with H2GCN gaining 8.77 points. On Amazon-Ratings, where node text is limited to short product titles, GAT improves by 6.78 points from fastText to MPNet, while GCN-sep changes by only 0.20 points. These results show that architectural performance is conditional on node representation: the same representation change can produce different magnitudes of performance gain across architectures, so architecture and representation cannot be treated as independent evaluation factors. A rank-correlation analysis on these two
benchmarks further shows that the relative ordering of architectures remains highly stable across representations, isolating differential sensitivity, rather than ranking instability, as the primary effect.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.DC
EfficientAgent: What Makes KV Cache Offloading Work for Concurrent Agents?
Kunming Shao, Jierun Chen, Jiangnan Yu et al. (+7 more)
LLM
agents resend their whole conversation on every turn, and most of it was already processed on the previous turn. Serving systems avoid recomputing it by caching its key-value (KV) state and, when GPU
memory runs out, by offloading that state to host
memory. For
agents, offloading gives inconsistent results: on the same coding-
agent workload it speeds up one deployment, slows down another, and changes nothing on a third, even where loading a
token back is several times cheaper than recomputing it. The reason is that cached state must survive until it is used again. While one
agent waits for its tool, the server processes the contexts of all other
agents, so an
agent's prefix is reused only if the host tier holds the reusable context of the whole
agent pool, which we call the reuse working set. A smaller tier keeps writing state that is evicted before anyone reads it. We present EfficientAgent, which sizes and manages the host tier by this working set. A stack-distance model estimates the working set from
agent histories to size the host tier; its predictions, made before the experiments, located the capacity at which offloading starts to pay. When the tier is too small, a runtime policy stops writing large refills of evicted context and keeps extending prefixes that are still cached; when the tier is large enough, it writes everything. On SWE-bench Verified coding
agents, a host tier sized to the estimated working set cuts recomputed
prompt tokens by 93% and end-to-end time by 39%. With a small fixed tier, the policy cuts recomputation by 35%; with a large tier, it avoids the 4.3-fold increase caused by always filtering writes. Across three GPU types and two models, offloading pays off when the GPU has little compute per byte of host bandwidth and the host tier holds the working set. Code is available at https://github.com/KunmingSHAO/efficientagent_release.
Read the full abstract ▾
All authors Kunming Shao, Jierun Chen, Jiangnan Yu, Xiao-Hui Li, Chaofan Tao, Yanli Wang, Huanxin Lin, Kwang-Ting Cheng, Chi Ying Tsui, Haoli Bai
Machine Learning
Marginal Calibration Masks Generator Differences in Held-Out Evaluation of Multivariate Time Series, with an Application to ERA5 Weather Sequences
Mustafa Özaytaç and Özge Karadağ Ataş
Generative models for multivariate time series are routinely evaluated with pooled distributional metrics computed after marginal calibration. We show that, under a held-out protocol, the calibrated Kolmogorov-Smirnov (KS) distance differs from a real-versus-real floor by at most the distance of the calibrated sample to the training block, so that it carries little information about the generator. We audit a three-discriminator GAN with evolutionary weight adaptation (StatD2GAN) against re-implemented RCGAN and TimeGAN baselines on ERA5 weather sequences under a pre-specified protocol: the final two calendar years are held out behind a 168 hour embargo, calibration is fitted on the training block only, and the decision rules were frozen in advance. Evidence comes from 25 matched (location, seed) pairs across five locations, tested with Holm-corrected Wilcoxon signed-rank tests. First, isotonic calibration drives KS to a median of 1.019 times a per-location noise-and-shift floor, and neither baseline is distinguishable from StatD2GAN on it. Second, the sorted-representation discriminator is the only component whose removal significantly degrades cross-variable dependence (Kendall tau MAE +0.116, Holm p = 0.002); a rank-transformed variant also degrades it, consistent with
quantile supervision rather than copula matching. Third, calibration raises physical constraint violations; projection removes them, but the largest marginal change (0.0110, Lhasa) exceeds the pre-specified tolerance in three locations. Fourth, the re-implemented TimeGAN is significantly better than StatD2GAN on Kendall tau and ACF MAE, RCGAN is worse on Kendall tau MAE and constraint violations, and the pre-specified collapse criterion was not met. We recommend floor-referenced marginal evaluation, matched-pair testing, and dependence and sequence-level metrics.
Read the full abstract ▾
Machine Learning
Oracle-Efficient Online Classification with Stochastic Inputs and Adversarial Outputs
Gon Buzaglo and Elad Hazan
We consider binary prediction with i.i.d. contexts from an unknown distribution and adaptively chosen losses. We show that a simple Follow-the-Perturbed-Leader
algorithm using a Gaussian perturbation for each observed context achieves $\widetilde O(\sqrt{T\log N})$ regret for a class of $N$ experts, while requiring one optimization-oracle call per round and no explicit enumeration of the class. For an infinite hypothesis class $\mathcal H$, the same
algorithm achieves $\widetilde O(\sqrt{T\operatorname{VC}(\mathcal H)})$ regret. This resolves an open problem posed by Lazaric and Munos (2012), showing that hybrid classification is computationally as easy as statistical learning. As an application, we reduce the problem of contextual bandits with $K$ actions to classification through uniform exploration, achieving $\widetilde O(K^{2/3}T^{2/3}(\log N)^{1/3})$ regret. This matches the best known dependence on the horizon while removing the context-distribution access required by prior oracle-efficient methods.
Read the full abstract ▾
Machine Learning · Cross-listed from cs.CL
Positions Are Not Facts: The Mismatch Between KV Caches and Memory
Changhai Zhou, Yuhua Zhou, Shiyang Zhang et al. (+5 more)
When a fact changes, how should a
language model update the history stored in its key-value (KV) cache? Hiding the old record is cheap, but it may still contain needed details or answer questions about the past. We compare hiding whole records, hiding only replaced values, and deleting old text and recomputing the cache. In a controlled quantity task, masking makes all eight models prefer the new value more strongly, yet six lose complete answers through unit errors or failure to stop; keeping the unit preserves all current answers. Later states also retain useful information from earlier records: on multi-hop updates, rebuilding these states at unchanged positions lowers historical accuracy by 20-41 percentage points, whereas moving the existing states has little effect. Keeping object dependencies and unchanged revision passages prevents many losses. Recognition is a separate challenge. Learned readouts recover distinctions missed by fixed cache similarities on synthetic record pairs. On natural text, text-detector-selected masks show no clear advantage over random masks at the same rate in 14 same-model detector-generator comparisons. Query-dependent access can avoid some losses, with additional storage or access costs. These findings identify what must be preserved beyond the replaced value when using a KV cache as updatable
memory.
Read the full abstract ▾
All authors Changhai Zhou, Yuhua Zhou, Shiyang Zhang, Jun Gao, Zhen Li, Hua Wu, Hanchao Yu, Haifeng Wang
Machine Learning · Cross-listed from eess.AS
YuE2: Unifying Symbolic and Audio Music Generation at Frontier Quality
Ruibin Yuan, Jiahao Pan, Junyan Jiang et al. (+32 more)
Symbolic models make melody, harmony, rhythm, and form explicit but typically stop before a finished recording; audio models produce complete songs while leaving composition implicit. We introduce YuE2, which unifies symbolic and audio music generation at frontier quality through symbolic planning. A single AR-NAR Mixture-of-
Transformers (MoT) first writes a readable score specifying melody and harmony, expands it into semantic music
tokens, and realizes it as full-song audio. In comparisons using the same checkpoint, experts prefer symbolic planning for overall quality and musicality, with 49.3% of overall preferences versus 34.6% without planning. Experts also favor the unified model over a separate
language model and
diffusion Transformer. On WildSongBench, YuE2 scores 6.73 on SongBench Global Avg, exceeding all evaluated public baselines. Selecting from eight candidates (best-of-8), YuE2 reaches 6.96, the highest observed mean among all evaluated systems. Expert listening further establishes its competitiveness with proprietary song generators, favoring best-of-8 over Suno v4.5 and yielding nearly balanced preferences against Suno v5. To learn this generation process from recordings without aligned scores, we introduce MERT2 and SheetSage2 to supply semantic and symbolic supervision. MERT2 sets a new state of the art in music representation learning, surpassing previous best results on 14 of 15 MARBLE metrics; SheetSage2 leads 12 of 15
benchmark-metric pairs in our lead-sheet transcription comparison. The same checkpoint follows score edits while largely preserving unedited musical content and generates zero-shot covers without cover-specific training. Its readable score also enables agentic music editing, with external
language models translating user feedback into revisions of the composition.
Read the full abstract ▾
All authors Ruibin Yuan, Jiahao Pan, Junyan Jiang, Zhiyue Wu, Ziya Zhou, Jiankai Sun, Yizhi Li, Ge Zhang, Yicheng Gu, Zeyue Tian, Junyu Dai, Hanfeng Lin, Kai Li, Shangda Wu, Xuanjie Liu, Jiaming Wang, Zihan Liu, Yue Wang, Yinghao Ma, Hanzhi Yin, Kangrui Chen, Xinyue Zhang, Ziyang Ma, Mengqi Liao, Hejia Zhao, Guowei Huang, Chao Yan, Lei Ke, Jianwei Yu, Bei Liu, Joe Guo, Liumeng Xue, Gus Xia, Wei Xue, Yike Guo
Machine Learning
Collaborative Synthetic Data for Privacy-Preserving Financial Fraud Detection Across Organizational Silos
Simeon Allmendinger, Domenique Zipperling, Burhanettin Bahadir Kibar et al. (+1 more)
Organizations seek analytical value from AI, yet relevant data are often fragmented across organizations and constrained by
privacy. This is acute in financial fraud detection, where rare fraud cases and imbalanced local
datasets limit decision-relevant analytics.
Federated learning enables collaboration without direct data sharing but does not resolve minority-class scarcity. Synthetic data generation can help, yet lightweight methods are interpolation-bound, while generative models require substantial data and computation. Existing collaborative generative approaches often rely on
federated learning, imposing considerable organization-side training burdens. In this paper, we examine CollaFuse as a collaborative
diffusion-based alternative for fraud detection and evaluate it across five fraud
datasets. Compared with classical oversampling, local generative baselines, and centralized
diffusion benchmarks, CollaFuse does not achieve the highest local fidelity but improves downstream fraud detection more consistently across most
datasets. These findings suggest that synthetic data create analytical value less through local realism than through transferable cross-organizational structure.
Read the full abstract ▾
All authors Simeon Allmendinger, Domenique Zipperling, Burhanettin Bahadir Kibar, Niklas Kühl
Machine Learning
ResDiffFRG: Residual Diffusion for Multiple Appropriate Facial Reaction Generation
Shizhe Liu, Jiayan Gu, Xiangyu Kong et al. (+1 more)
In dyadic human speaker-listener conversations, the listener's facial reactions allows the speaker to accurately perceive the listener's emotional states. Since human facial reactions are non-deterministic, the ability to generate multiple appropriate human-like facial reactions is crucial for realistic human-
agent interactions. Although
diffusion models are naturally suited to such one-to-many generation, existing
diffusion-based Multiple Appropriate Facial Reaction Generation (MAFRG) methods attempt to denoise random Gaussian initialisations directly into multiple appropriate facial reactions (AFRs). These random initialisations are usually not well-aligned with the target listener facial reaction, which requires complex denoising trajectories from these initialisations, and subsequently creates substantial opportunities for deviations away from the range of trajectories leading to appropriate AFRs. Given the inherent mimicry between the human listener's and speaker's facial behaviours, we address the above denoising trajectory issue by leveraging this strong prior. Specifically, we propose ResDiffFRG, a novel
diffusion-based MAFRG framework that explicitly anchors the
diffusion process to the speaker behaviour by defining its
diffusion target as the residual between the speaker anchor and an AFR. The denoiser only needs to model the comparatively small, reaction-specific residual needed to transform this anchor into an AFR, rather than reconstructing the complete reaction from an unstructured state. Extensive experiments show that ResDiffFRG achieves large improvements in correlation-based appropriateness over existing methods. Our denoising trajectory analysis showed that even at the start of the denoising trajectory, ResDiffFRG already achieves a higher facial-reaction correlation score than the Gaussian
Diffusion baseline does after completing 60% of its denoising trajectory.
Read the full abstract ▾
All authors Shizhe Liu, Jiayan Gu, Xiangyu Kong, Siyang Song
Machine Learning
Task-Aware Discretization of Differentiable Logic Gate Networks
Thore Gerlach
Differentiable logic gate networks (DLGNs) enable gradient-based training of highly efficient Boolean networks by relaxing discrete logic gates during training and discretizing them for inference. Standard approaches make this discretization decision locally, typically through argmax selection and confidence- or
entropy-based convergence criteria. We show that local discretization can be task-suboptimal even for globally optimal relaxed solutions, with high gate confidence providing no general guarantee, and derive bounds relating task-aware gate selection to tractable interventions in the relaxed network. Motivated by these results, we study first-order downstream task information for progressive discretization and characterize when this local approximation is reliable. Experiments on convolutional DLGNs reveal a strong locality dependence: first-order scores become unreliable when directly optimized over nonlocal interventions, but accurately assess local argmax decisions for progressive freezing.
Read the full abstract ▾
Machine Learning
PQ-HSA: Reusing Product-Quantized Scores for Hybrid Sparse-Approximate Attention
Kunming Shao, Jierun Chen, Yanli Wang et al. (+4 more)
At each decoding step a
language model attends over the key-value (KV) cache of every earlier
token, so at long context the
attention call is bounded by
memory bandwidth. Sparse
attention reads only a subset of keys chosen by a cheap score estimate, and most methods give the unread
tokens zero weight. The output then draws on only a small fraction of the KV cache, and accuracy drops at small budgets, most on tasks that aggregate information across the context. An inverted-file product-quantization (IVF-PQ) index over the cached keys computes an approximate score for every indexed
token in order to rank them; after ranking, those scores approximate the
attention logits of the
tokens left out. PQ-HSA (hybrid sparse-approximate
attention) attends the selected
tokens with their original keys and values, and the unselected
tokens, the background, enter the same softmax through those scores, summed per inverted list and multiplied by the list's mean value. At 128K and a 1-2%
retrieval budget, PQ-HSA is more accurate than Quest and SnapKV on Llama-3.1-8B and Qwen3-30B-A3B and stays close to full
attention in macro accuracy; with the same selector, the background term raises macro accuracy on the 8B model from 0.71 to 0.83. In the same 128K setting, inside vLLM on one NVIDIA H20, the decode
attention call runs 1.6x faster than the FlashAttention-3 kernel; the speedup grows with context length, and a cost model fitted on 8B to 30B models gives the context length at which it begins. A vLLM plugin runs PQ-HSA on two engine versions without changes to the engine source; code is available at https://github.com/KunmingSHAO/pqhsa_release.
Read the full abstract ▾
All authors Kunming Shao, Jierun Chen, Yanli Wang, Ruoyu Wang, Haoli Bai, Kwang-Ting Cheng, Chi Ying Tsui