arXiv subjects

Ming Li

Publications and source records attributed to Ming Li.

At least 19 recordsLinked to original sources

arXiv2026

From Voiceprints to User Entities: Online Speaker-Memory Construction for Multi-User Spoken Agents

Personalized spoken agents must identify speakers before accessing user-specific memory, yet memory benchmarks typically assume correctly attributed histories. We argue that voiceprints are observations of identity, not identities themselves, and formulate online speaker-memory construction to maintain persistent user entities as acoustic evidence evolves. We introduce EvoSpeakerMem, a training-free layer for causal speaker routing, selective acoustic-state updates, and identity-partitioned memory writes and retrieval through a replaceable text-memory backend. We also construct LoCoMo-ID, a spoken-memory benchmark with hidden speaker identities, speaker-sensitive queries, simulated recording-condition shifts, and four-speaker identity-confusion tests. On longitudinal SMIIP-TV, genuine-probe rejection under a frozen near-enrollment threshold rises from 29.8% to 55.6% over approximately one year. On LoCoMo-ID speaker-sensitive queries, gated updating improves identity-gated QA over fixed representations by 3.86 and 11.62 percentage points under clean and shifted conditions, respectively, with no wrong-user authorizations. The advantage persists in the four-speaker test. Yet reducing fragmentation and improving QA can coexist with incorrect cross-user links, revealing a fragmentation--contamination trade-off that QA alone cannot capture.

cs.SD↗

arXiv2026

A Lightweight Dual-Branch Visual Front-End for Efficient Audio-Visual Target Speaker Extraction

Audio-Visual Target Speaker Extraction (AVTSE) uses face-related visual cues to extract a target speaker from a speech mixture. In many AVTSE systems, visual features are obtained from convolutional image encoders, and their computation may become a non-negligible part of the overall system when lightweight separation backbones are used. This work proposes a dual-branch visual front-end that reorganizes visual information according to its spatial characteristics. A compact lip region of interest is used to model local articulation cues through convolutional processing, while facial landmark coordinates are used to represent global geometric motion with a lightweight temporal convolutional network. The two representations are integrated into IIANet and SpatialNet with limited fusion overhead. Experiments on LRS3-2mix under a unified reduced training protocol show that the proposed method achieves a performance-computation trade-off close to the lightweight image-based baseline. Compared with the ResNet18-based visual encoder, it reduces the module-level visual encoding computation by more than 90% across both evaluated separation backbones, with a moderate performance gap. This comparison excludes face detection and landmark estimation and therefore applies to systems where landmark coordinates are already available from preprocessing. Demos are available at https://anonymous.4open.science/r/Lightweight-Visual-Frontend-E46E.

eess.AS↗

arXiv2026

SLT 2026 REAL-TSE Challenge: Real-world Target Speaker Extraction from Conversational Recordings

We introduce the REAL-TSE Challenge, an IEEE SLT 2026 satellite challenge on target speaker extraction~(TSE) from real conversational recordings. Given a multi-speaker mixture and one or more enrollment utterances from a target speaker, participating systems must recover only the target speech. Unlike simulated read-speech benchmarks, REAL-TSE evaluates Mandarin and English recordings that contain natural overlap, reverberation, noise, channel mismatch, and conversational dynamics. The challenge defines two complementary tracks: an Online track for low-latency streaming extraction and an Offline track for full-context processing. Systems are evaluated with Token Error Rate (TER), Speaker Similarity (SpkSim), DNSMOS, and target-speaker activity F1. This overview paper describes the task definition, datasets, baselines, evaluation protocol, submitted systems, condition-wise findings, and lessons for future real-world TSE benchmarks.

eess.AS↗

arXiv2026

LongSpark: Efficient speculative decoding with a fixed-cost parallel drafter

Speculative decoding accelerates autoregressive inference by verifying multiple draft tokens in a single target forward pass. However, as the context grows, existing state-of-the-art drafters become increasingly expensive, eroding the very efficiency advantage they are designed to provide. We argue that this scaling is unnecessary. A standalone language model must grow with its prefix because it is solely responsible for every token it produces. A drafter, by contrast, only proposes candidates; the target catches and corrects every error before any token is committed. The drafter's decoding cost can therefore be made entirely independent of the prefix length. We introduce LongSpark, a block-diffusion drafter that achieves this by extracting fixed-size, multiscale views from the target's verification pass, thereby eliminating the need for a growing persistent state. Extensive evaluations demonstrate that LongSpark achieves state-of-the-art end-to-end efficiency across multiple model scales and realistic serving conditions. Notably, it delivers the lowest time-per-output-token on long-context tasks while reducing the drafter's context state by several orders of magnitude.

cs.LG↗

arXiv2026

S2SND-v2: Unified Online and Offline Speaker Diarization with Speaker-Adapted w2v-BERT 2.0

This paper presents S2SND-v2, a unified online and offline speaker diarization framework built on S2SND that improves acoustic feature extraction and the use of the resulting representations. For acoustic feature extraction, we introduce a speaker-adapted w2v-BERT 2.0 front end and compare full and pruned construction routes. To better use these features, overlap-aware activity-query masking improves speaker embedding extraction under overlap, while late-speaker training aligns unknown-speaker training with online inference conditions. Combining these improvements, the pruned system achieves competitive average diarization error rates of 14.52% online and 12.03% offline across eight test sets.

cs.SD↗

arXiv2026

One Sentence, One Drama: Personalized Short-Form Drama Generation via Multi-Agent Systems

Existing approaches for digital short-drama production typically rely on one-shot LLM generated scripts and loosely coupled pipelines, which fail to satisfy three key requirements of short-drama generation: (1) narrative pacing, resulting in weak hooks, insufficient escalation, and unattractive endings; (2) spatial consistency, leading to drifting scene layouts and inconsistent character positions across clips; and (3) production-level quality control, requiring extensive manual review and correction across script and visual stages. We present One Sentence, One Drama, a hierarchical multi-agent framework that transforms a user's single-sentence idea into a fully produced short drama through structured intermediate modules and iterative refinement. Our approach is built upon three key components: (1) a multi-agent debate-based story generation module that enforces short-drama pacing and narrative coherence; (2) a 3D-grounded first-frame generation mechanism that establishes a shared spatial reference for consistent character positioning and scene layout across clips; and (3) multi-stage reviewer loops that perform comprehensive error detection and targeted revision across script, visual, and video generation stages. We also introduce scene-level BGM matching and scene transition planning to improve the audience's immersive experience. To systematically evaluate this task, we introduce Short-Drama-Bench, a benchmark that extends standard video quality metrics with short-drama-specific criteria. Experimental results demonstrate that our method significantly outperforms existing pipelines in narrative quality, cross-clip consistency, and overall viewing experience.

cs.CV↗

arXiv2026

ASR-Protected Fusion and Reference-Anchored Policy Optimization for Joint ASR and Word-Level Speech Anti-Spoofing

Existing anti-spoofing systems report either an utterance-level score or a time-based spoof interval: the former cannot show which words are spoofed, while the latter does not directly match the transcript and must be post-aligned with words, so alignment errors can lead to wrong word-level labels. Standard ASR model outputs only text; spoof information in the audio is lost after recognition and is not passed with the transcript. Yet changing only a few words can change the meaning of a sentence, leaving downstream systems to use incorrect text without knowing which words are spoofed. We therefore treat partial anti-spoofing as tagged ASR: the model transcribes speech and marks spoofed words at the same time, directly linking spoof labels to output words. We adopt the audio LLM and use ASR-Protected Subspace Gated Fusion (APSGF) to fuse spoof cues with ASR features. Spoof cues are added only outside an ASR-protected subspace, so they do not disturb recognition features. After cross-entropy training, we fine-tune the model with Reference-Anchored Policy Optimization (RAPO). Since the SFT policy is already concentrated, multiple samples from the same input tend to receive similar rewards, resulting in weak group-wise advantages. RAPO instead generates a single sample with low sampling randomness and anchors its advantage to the reward of a reference model. This allows the model to reinforce responses that outperform the reference while suppressing those that underperform it. Experiments on five datasets demonstrate the effectiveness of our approach in jointly detecting spoofed words and preserving ASR accuracy. The code is publicly available.

cs.SD↗

arXiv2026

PhysCreator: Creating Physically-Attributed 3D Assets at Will

Significant advancements in 3D asset creation have enabled high-fidelity 3D generation from images. However, most existing representations remain inherently static, lacking the ability to model physically plausible behaviors under dynamic interactions. We present \textbf{PhysCreator}, a unified framework that directly generates simulation-ready 3D representations with spatially varying physical properties from images. Unlike prior approaches that treat physical attributes as simple post-hoc predictions or rely on expensive per-instance optimization, we reformulate physical property estimation within a unified 3D representation learning framework. We introduce a voxel-aligned latent space by lifting image observations into a spatially structured 3D representation, in which geometry, appearance, and physical properties are jointly modeled and inferred in a feed-forward manner. To enable scalable learning without densely annotated physical ground truth, we leverage vision-language distillation to transfer material-aware semantic priors from powerful vision-language models into spatially grounded voxel-level supervision. Experiments demonstrate that our method achieves superior visual quality, physical plausibility, and generalization compared to prior approaches.

cs.CV↗

arXiv2026

Weizsäcker-Williams Gluon Helicity Distribution and Inclusive Dijet Production in Longitudinally Polarized Electron-Proton Collisions

It is well-known that the back-to-back (correlation) limit of inclusive quark-antiquark dijet production in unpolarized high energy electron-proton collisions can probe the Weizsäcker-Williams (WW) gluon transverse momentum-dependent distribution (TMD) at small $x$ \cite{Dominguez:2010xd, Dominguez:2011wm}. In this paper, we consider a helicity-dependent version of the same process: we study the double-spin asymmetry for inclusive quark-antiquark dijet production in longitudinally polarized electron-proton scattering at high energies. We show that in the back-to-back limit this process probes the WW gluon helicity TMD. Furthermore, we derive the small-$x$ evolution equation for the operator related to the WW gluon helicity distribution. We find that in the double-logarithmic approximation and in the large-$N_c$ limit, the small-$x$ asymptotics of the WW gluon helicity distribution is governed by exactly the same evolution equation as that for the dipole gluon helicity distribution. The longitudinal double-spin asymmetry for inclusive dijet production in the longitudinally polarized electron-proton collisions can thus test the small-$x$ helicity evolution equations and facilitate constraining the initial conditions for phenomenology based on these equations.

hep-ph↗

arXiv2026

Thermodynamic formalism for singular flows

We prove that $C^\infty$ smooth three-dimensional flows with positive topological entropy admit only finitely many ergodic measures of maximal entropy, even in the presence of singularities (zero-velocity points). Furthermore, for flows in a $C^\infty$-open and dense subset of the space of such three-dimensional flows, every ergodic measure of maximal entropy is rapidly mixing. To prove this result, we construct a novel symbolic coding framework for singular flows, which serves as the fundamental technical tool. We then introduce the property of strong positive recurrence (SPR) for singular flows, and prove that SPR flows can be coded by suspension flows over SPR countable discrete symbolic systems. Under appropriate conditions, we further establish that such suspension flows are rapidly mixing. These frameworks can be further generalized to a broad class of flows, including star flows, and the results naturally extend to general equilibrium states.

math.DS↗

arXiv2026

Orbital angular momentum at small $x$ in the large $N_c\&N_f$ limit

We extend the small-$x$ analysis of the quark and gluon orbital angular momentum (OAM) distributions in the proton from the large-$N_c$ limit considered in our earlier works to the large-$N_c\&N_f$ limit, in which the numbers of quark colors $N_c$ and flavors $N_f$ are large with their ratio held fixed. Working in the double-logarithmic approximation (DLA), summing powers of $α_s \ln^2(1/x)$ with $α_s$ the strong coupling and $x$ the proton's momentum fraction carried by a parton, we correct the small-$x$ operator expression for the quark OAM distribution suggested earlier in \cite{Kovchegov:2024wjs} and relate both the quark and gluon OAM distributions to the impact-parameter moments of the polarized dipole amplitudes. We derive the large-$N_c\&N_f$ evolution equations for these moment amplitudes in the DLA; these include a new equation for the moment amplitude governing the quark OAM. We then solve these evolution equations numerically together with the helicity evolution for $N_f = 2,3,4,5,6$ and $N_c =3$. We find that, similar to the large-$N_c$ case, the OAM distributions share a common small-$x$ intercept with the helicity distributions, $L_{q+\bar{q}}(x,Q^2) \sim L_G(x,Q^2) \sim ΔΣ(x,Q^2) \sim ΔG(x,Q^2) \sim \left(\frac{1}{x} \right)^{α_h}$ with the intercept $α_h \approx {3.48 \sqrt{α_s N_c/2π}}$ for $N_f = N_c = 3$ (cf.~\cite{Borden:2025ehe}): this result, along with $α_h$ for other values of $N_f \neq 0$ that we studied, is smaller than the intercept of $3.66\sqrt{α_s N_c/2π}$ found in the large-$N_c$ limit. We also compute the ratios of the OAM distributions to the helicity parton distribution functions as $x\to 0$, obtaining $L_{q+\bar{q}}(x,Q^2)/ΔΣ(x,Q^2) = -1$ and $L_G(x,Q^2)/ΔG(x,Q^2) = -2$ at $Q^2=10\, \mathrm{GeV}^2$, with both ratios being independent of $N_f$.

hep-ph↗

arXiv2026

AgenticRoboRL: Bringing Coding-Agent Post-Training to Agentic Robotics

Agentic robotics systems combine planners with robot harnesses to perform complex manipulation tasks. Recent work improves these systems through better harnesses and external memory, but relies on powerful closed-source models as planners and does not improve the planners' own capabilities. In contrast, coding agents have improved substantially by post-training the model itself. We therefore investigate whether post-training can also improve robot planners. However, applying it across robot harnesses requires accommodating different execution mechanisms, including VLA skill invocation and control-code generation. To this end, we introduce AgenticRoboRL, a framework for post-training robot planners across different harnesses. The training recipe combines supervised fine-tuning with multi-turn reinforcement learning. To accommodate different execution mechanisms, harness-specific adapters connect recorded planner interactions with task outcomes, enabling a common RL training pipeline while keeping execution components fixed. On LIBERO-PRO, AgenticRoboRL raises the success rate of a 9B-parameter planner from 18.1% to 71.0% without external memory, approaching the 72.1% achieved by Codex (GPT-5.5) with external memory. These results demonstrate post-training as an effective approach for improving robot planners. We will release the framework and trained planner checkpoints.

cs.RO↗

arXiv2026

Evolving Intelligent Complex Systems via Intellicise Networks: Architecture, Technologies, and Pathways

Future engineering infrastructures are evolving into large-scale, open, heterogeneous, and wirelessly interconnected complex systems. These systems present significant challenges in optimizing network resource utilization, managing high-dimensional information spaces, and accommodating diverse business requirements. Intellicise networks, characterized by Intent-driven operation, semantic-native capability, and distributed intelligence, offer a promising paradigm for enabling such intelligent complex systems. We provide a systematic exploration of future intelligent complex systems from the perspective of intellicise networks. Specifically, we propose a cross-domain intelligent communication network architecture based on intellicise networks, grounded in information theory, systems theory, game theory, and cybernetics. The architecture comprises a cross-layer organizational framework, multi-functional planes, and novel information flows. The cross-layer framework defines the vertical evolution from perception and cognition to decision, while the control, user, data, computation, intelligence, and security planes deliver horizontal intellicise capabilities. Moreover, data, knowledge, model, and task flows interconnect the various layers and planes, forming a closed-loop process that derives simplicity from high-level intelligene while concurrently pursuing enhanced. Building on this architecture, we review key enabling technologies, tracing their evolution from semantic extraction to intent understanding, from heterogeneous resource integration to self-configuration and self-optimization, from generative artificial intelligence (AI) to agentic AI, and from embodied AI to symbodied AI. Additionally, we present a case study on intellicise networks for embodied agent communications and discuss representative applications and services for intelligent complex systems.

eess.SP↗

arXiv2026

Hallucinations of Large Language Models in Cross-Language Code Translation: Phenomena and Mitigation

Code translation converts source code from one programming language to another while preserving its functionality. Although Large Language Models (LLMs) have improved code translation, their outputs still contain hallucinations: generated code that conflicts with the source program, violates target-language facts, or introduces contextually inconsistent dependencies. These failures are especially harmful in translation because small semantic deviations can silently change program behavior. Existing hallucination studies mainly focus on natural-language generation or code generation, leaving the forms, distribution, and mitigation of hallucinations in cross-language code translation insufficiently characterized. This paper presents a systematic empirical study of LLM hallucinations in code translation. Across three benchmark datasets, six language pairs, and seven LLMs, we construct a taxonomy with three major categories: Semantic Consistency Conflicts, Factual Knowledge Conflicts, and Project Context Deviations. We manually annotate 13,919 hallucination cases and analyze how hallucination types vary across models, datasets, and language directions. Guided by the taxonomy, we further propose HaluCodeTran, a mitigation framework that combines BM25-based retrieval of relevant translation examples with taxonomy-aware reflective refinement. The study provides an execution-tested characterization of code-translation hallucinations and a practical baseline for building more reliable LLM-based code migration tools.

cs.SE↗

arXiv2026

Beyond the Teacher: Understanding and Preserving Distillation Overshoot in Agentic On-Policy Distillation

On-policy distillation (OPD) trains agents on their own trajectories using supervision from a stronger teacher. We uncover a counterintuitive phenomenon in multi-turn agentic tasks: a smaller, initially weaker student can substantially outperform its teacher during training, only to regress as distillation continues. We term this phenomenon distillation overshoot. Through decision- and task-level analyses, we trace overshoot to two competing effects of OPD: repairing weak student decisions and eroding successful student choices that the teacher disfavors. Early in training, repair outpaces erosion, allowing the student to combine the teacher's corrections with its own strengths and surpass the teacher. As training continues, erosion overtakes repair, causing these gains to vanish. Motivated by this finding, we propose Success-Preserving On-Policy Distillation (SPOD), which selectively masks conflicting teacher supervision on successful student choices and introduces a soft probability floor to counteract their erosion, while retaining teacher guidance elsewhere. SPOD requires only binary episode success and model probabilities already available during OPD. Across two model families and three agentic benchmarks, SPOD consistently outperforms ordinary OPD and evaluated variants, improving final success by up to 15.2 percentage points and surpassing the teacher by up to 20.0 points. Moreover, SPOD's final checkpoints outperform even ordinary OPD's best intermediate checkpoints in all six settings, demonstrating that beyond-teacher gains can be not only preserved but further extended. Code: https://github.com/Jike338/SPOD

cs.LG↗

arXiv2026

Toward Collective Model Intelligence: Characterizing and Coordinating Models

Independently developed AI models accumulate diverse and complementary capabilities, but a growing model collection does not automatically realize Collective AI. Addressing this gap requires understanding model capabilities with limited evaluation, organizing complementary models to collaborate on diverse tasks, and executing model calls efficiently under resource constraints. We present a systematic approach to pursuing Collective Model Intelligence through three connected components. First, Model Eval constructs evidence-based capability profiles by combining query-level performance estimation with cross-domain prediction, reducing the need for exhaustive evaluation. Second, Mio (Models in One) organizes complementary model candidates and supports task-conditioned, cost-aware selection and collaboration as task requirements evolve. Third, Mos (Model Serving OS) separates model selection from execution and provides endpoint scheduling and shared model deployment and distribution. Experiments show lower capability-estimation errors under limited evaluation. Across nine public benchmarks, Mio 1.0 Lite achieves the highest average score and lowest average cost among compared methods, with per-request costs 73.7-97.5% lower than those of the best-scoring alternatives. A prewarmed shared cache accelerates model-file retrieval by up to 23.3 times. This work lays a foundation for Collective Model Intelligence by characterizing and coordinating the diverse capabilities of an expanding model universe.

cs.AI↗

arXiv2026

Mapping Dark Matter in Merging Clusters with Weak Lensing for CSST: Morphology and Filament Recovery

Weak gravitational lensing offers a direct probe of the projected dark matter distribution in cluster-scale structures and the surrounding cosmic web. We investigate the capability of the China Space Station Telescope (CSST) to recover non-spherical dark matter morphologies using mock reduced-shear maps generated from cosmological N-body simulations. We apply the deep-learning-based MIU2Net mass-reconstruction framework to CSST-like wide- and deep-field source densities, focusing on two representative systems: filament-connected halo pairs and elongated halos. MIU2Net accurately recovers halo centers, elongated mass distributions, and the filamentary structures connecting halo pairs in both survey configurations. Halo centers are localized within about 0.1 h^(-1) Mpc, and elongated-halo shapes and orientations are reliably recovered. Filament paths are stable against shape noise, with mean normalized instability parameters I_F, defined as the mean deviation from the true filament locus in units of R_200, decreasing from 0.041 to 0.037 in the low-redshift bin and from 0.051 to 0.045 in the high-redshift bin from the wide to deep field. The recovered filament shapes remain consistent with the true maps at the percent level. Compared with conventional Kaiser-Squires reconstruction, MIU2Net better preserves low-contrast anisotropic structures while suppressing small-scale noise. These results demonstrate that CSST weak-lensing data, combined with machine-learning-based mass reconstruction, will enable statistical studies of cluster assembly, anisotropic matter accretion, and the connection between massive halos and the filamentary cosmic web.

astro-ph.CO↗

arXiv2026

Failure Recovery via Belief Updates and World Modeling for Long-Horizon Agents

Long-horizon agents often struggle to recover from failures whose causes lie several steps earlier, beyond what local reflection or backtracking can resolve. We introduce BELIEFWEAVER, a framework that combines persistent evidence across attempts with prospective reasoning over repair directions. It maintains a relational belief state over frontiers, attempts, observations, and checkpoints. Execution evidence updates related frontier priorities across attempts and rollback, while diagnostic evidence gates execution. An implicit world model (IWM) evaluates repair hypotheses, predicts progress, and identifies relations among frontiers to guide exploration. We further distill these recurring IWM reasoning components into a 9B student through supervised fine-tuning and on-policy distillation (OPD). Across three terminal-based benchmarks, BELIEFWEAVER improves aggregate final score by 5.1 percentage points over native Terminus-2 continuation, with positive trends in exploration efficiency. The distilled 9B student improves final score by 9.8 percentage points over the reported native baseline while reducing estimated teacher API cost by 54.8% relative to the matched teacher. These results show that persistent evidence and prospective reasoning can improve recovery in long-horizon agent tasks while substantially reducing reliance on teacher-model inference.

cs.LG↗