arXiv subjects

Junjie Li

Publications and source records attributed to Junjie Li.

At least 19 recordsLinked to original sources

arXiv2026

Uncertainty-Guided Multimodal Consensus for Multimodal Sentiment Analysis

Multimodal sentiment analysis (MSA) has made considerable progress in recent years. Existing MSA methods mainly learn shared multimodal representations or train unimodal predictors with a common multimodal label. However, assigning the shared label to every modality can obscure modality-specific sentiment, while direct fusion may overemphasize unreliable predictions when modalities disagree. To address these issues, we propose Uncertainty-Guided Multimodal Consensus (UMC), which first recovers continuous unimodal pseudo-labels and then forms a reliability-aware consensus across modalities. Our method consists of two key components: i) Continuous Prototype-Guided Label Recovery (CPLR), which combines heteroscedastic predictions with ordered sentiment prototypes and uses confidence and label conflict to refine continuous unimodal pseudo-labels, and ii) Reliability and Precision Consensus (RPC), which estimates sample-dependent modality reliability from predictive statistics and cross-modal disagreement and combines it with predictive precision through a closed-form consensus solver. Extensive experiments conducted on four publicly available MSA datasets demonstrate the effectiveness and superiority of our proposed model.

cs.CV↗

arXiv2026

MMA-SafetyBench: A Benchmark for Multimodal Agent Safety Evaluation

Multimodal agents increasingly rely on visual interfaces to perceive state and select actions, creating an attack surface not captured by text-only safety evaluations. We study Visual Semantic Hijacking, in which deceptive but semantically meaningful interface elements, such as forged system alerts, cause an agent to depart from its authorized task. We propose a trajectory-guided procedure that identifies candidate injection points and inserts reproducible visual payloads into the observation stream. Using this procedure, we construct MMA-SafetyBench, which contains 1,083 adversarial trajectories across web automation, desktop GUI control, mobile navigation, document understanding, and video processing. Across nine evaluated models, Attack Success Rate (ASR) reaches 87.96%, while Semantic Compromise Rate (SCR) reaches 98.09%. A 1--3--5 scoring scheme distinguishes target-matching outputs, semantic disruption without a target match, and successful preservation of the benign task. Controls covering all nine models and five domains show that attacks persist under neutral prompts. A safety-priority prompt lowers ASR without eliminating the failure mode.

cs.CR↗

arXiv2026

LiFi: LiDAR Generation from Multi-View Images via Geometric and Semantic Collaborative Guidance

Generating LiDAR point clouds from multi-view camera images is a valuable task with applications in controllable data synthesis, cross-modal simulation and various perception tasks. However, most existing methods are not specifically designed for this task, and the quality of their generated LiDAR remains limited. Moreover, these methods typically focus only on semantic cues in images or generate LiDAR from a monocular camera image. In this work, we introduce LiFi (LiDAR From images), a dedicated framework for generating visually and geometrically aligned LiDAR scenes from multi-view camera images. LiFi guides the latent diffusion process through two complementary branches: geometry and semantics. The geometry branch constructs scene features from images through a reverse sampling algorithm and an uncertainty-aware geometric encoder combined with DepthAnything3, whereas the semantic branch provides high-level semantic representations via a cross-view interaction module. We further introduce a dual-branch balanced classifier-free guidance (CFG) strategy, which enhances the model's conditional generation capability while preserving the independent learning and collaborative controllability of the two branches. Extensive experiments demonstrate that LiFi outperforms state-of-the-art methods in generating high-fidelity LiDAR scenes. The code will be publicly available at: https://github.com/szzhou88/LiFi.

cs.CV↗

arXiv2026

Strong convergence rate of the implicit midpoint method for stochastic KdV equations

We establish a strong convergence rate for the implicit midpoint method applied to the periodic stochastic Korteweg-de Vries equation with additive noise, addressing a question that has remained open since the work of Debussche and Printems. To control nonlinear error growth, we obtain exponential integrability estimates for the exact and comparison solutions using the coercivity of a classical KdV invariant. We then derive local error bounds from estimates for the comparison solutions and control error accumulation through an $L^2$ error propagation identity that isolates a martingale increment. Under $H_0^{9/2}$ regularity assumptions on the noise and the deterministic initial data, we prove that, on every fixed finite time interval, the maximum $L^2(\mathbb{T})$ error at the time points is $O(h)$ in $L^p(Ω)$ for every $1\le p<\infty$. Notably, the exponential integrability and the convergence rate hold for every damping coefficient $λ\ge0$, including the undamped case $λ=0$. Numerical experiments confirm the predicted first-order convergence.

math.NA↗

arXiv2026

Bulkhead: Automated Semantic Detection and Remediation of Container Escape Vulnerabilities

Container filesystem isolation depends on safely resolving container-controlled paths during privileged host-container interactions. As third-party plugins and agent sandboxes introduce more interactions, unsafe path resolution can redirect host operations across isolation boundaries, expanding the attack surface for path traversal vulnerabilities. These logic flaws are difficult to mitigate: kernel defenses require intrusive redesign, static rules miss cross-boundary semantics, and manual auditing cannot scale. Large Language Model (LLM) based methods can detect syntax flaws through stable code patterns and identify logic flaws in domains such as smart contracts by comparing equivalent implementations. Container components, however, provide neither form of regularity: their heterogeneous interfaces operate under trust boundaries, filesystem views and interaction orders, making the security invariant difficult to find. We present Bulkhead, an automated framework for detection and remediation. Bulkhead summarizes historical CVEs into structured semantic patterns, correlates documentation with entry and effect side code evidence, and recovers source-to-sink execution contexts. Grounded patterns guide LLM reasoning, while isolated Proof-of-Concept (PoC) execution validates findings. For remediation, Bulkhead generates repository-specific patches, checks that they build, and iteratively verifies path safety and functionality using pattern derived model checking templates and counterexamples before runtime validation. Across container repositories, Bulkhead discovered 20 zero-day vulnerabilities, accounting for 43% of the total number discovered over the past decade. Among them, 3 have been assigned CVEs and the verified patch suggestions have been submitted to the upstream.

cs.CR↗

arXiv2026

GateSight: Recurrent Memory Anticipates Attention Memory for Query-Agnostic KV Eviction in Hybrid LLMs

Hybrid language models interleave Gated DeltaNet (GDN) layers, whose recurrent memory has a fixed size, with full-attention layers, whose key-value (KV) cache grows with the context and dominates serving memory. In this paper, we ask whether the recurrent memory already knows what the attention memory will need. We first show that which GDN heads of Qwen3.8-27B keep long memories is fixed by the weights rather than by the document. Next, we define the future utility of a KV entry by how much removing it changes the attention output over the model's own continuations, and find that the write gate, decay and write magnitude that GDN layers emit during prefill anticipate this utility before any query is seen. Because the relation is spread across many GDN heads and layers, and a linear map recovers its trend but less of its high-utility tail than a nonlinear one, we propose GateSight, which learns it with lightweight per-head multilayer perceptrons that read these signals and no attention score. Their predicted utility decides which entries each attention layer keeps, and how much each layer keeps is set by the model's native attention output gate, whose mean differs by more than an order of magnitude across layers. Finally, at serving time, GateSight scores the cache once after prefill and physically releases discarded KV entries. Experimental results on SCBench and RULER-4096 with Qwen3.8-27B and, with the same recipe, the sparse-attention model Qwen3.8-Flash-Next, up to the 262K-token limit of the models, show that with a tenth of the attention KV cache, GateSight is the most accurate of five query-agnostic methods on three of the four model-benchmark pairs and leads KVzip on all four, by 9.5 to 47.2 points. Notably, it scores a full-length context with negligible overhead relative to prefill.

cs.AI↗

arXiv2026

V-ICAL Bench: Evaluating Video In-Context Learning for Multimodal Agents in Interactive Environments

While In-Context Learning (ICL) enables models to adapt from exemplars without parameter updates, multimodal ICL remains largely underexplored, particularly regarding video demonstrations in interactive environments. For multimodal agents, learning from videos presents unique challenges: they must translate in-context demonstrations into executable policies, ground these policies in novel visual states, and iteratively refine actions based on environmental feedback. We introduce V-ICAL, a novel benchmark designed to evaluate video-based ICL for multimodal agents. Comprising 342 interactive tasks across 37 environments, V-ICAL utilizes human-curated demonstration videos as task-specific behavioral exemplars, evaluating agents through sustained interaction from a target initialization. The benchmark seamlessly connects in-context knowledge induction with core agentic capabilities, including state grounding, temporal memory, planning, and adaptation in dynamic environments. Extensive evaluations across 19 state-of-the-art multimodal agents reveal significant limitations: the best-performing model, Seed-2.1-Pro, achieves a score of only 54.4/100, while other leading models (e.g., Gemini-3.1-Pro, GPT-5.6) fail to surpass 50, far below the human baseline of 83.6. Controlled comparisons further demonstrate that current agents struggle to reliably translate video exemplars into effective policies, failing to yield consistent performance gains. Ultimately, V-ICAL exposes a critical gap in the ICL capabilities of multimodal agents, underscoring an urgent need for future research.

cs.CV↗

arXiv2026

TS-SP: Learning Speaker-Preserving Representations in Audio Large Language Models

Audio large language models (ALLMs) can understand speech content, yet their ability to use speaker identity for verification remains limited. We propose TS-SP (Two-Stage Speaker Preservation), a parameter-efficient framework for learning speaker-preserving representations and making them accessible to an ALLM's language-model component. We instantiate and evaluate TS-SP on Qwen2.5-Omni-7B. First, we adapt the audio encoder with speaker identity supervision. We then freeze the adapted encoder and train the language model to compare speakers. Both stages use low-rank adaptation (LoRA), keeping the pretrained base weights fixed. On Vox1-O, TS-SP reduces the equal error rate (EER) from 7.01\% for the Paired Loss Adaptation Baseline to 4.37\%. EER remains within 4.31--4.79\% under unseen prompts. Cross-domain evaluation on CN-Celeb yields a similar EER to the baseline, but lower accuracy at the native decision threshold. These findings support two-stage adaptation for improving speaker verification on the evaluated backbone.

cs.SD↗

arXiv2026

The ISCSLP 2026 Real-World Audio-Visual Speech Enhancement Challenge: A Benchmark with Natural Mixtures and Degraded Video

Audio-visual speech enhancement (AVSE) uses a target speaker's visible articulation to recover that speaker's voice from overlapped speech. Yet most evaluation protocols rely on synthetic mixtures and reliable video. The ISCSLP 2026 Real-World AVSE Challenge addresses both gaps. Track~1 combines naturally recorded two-talker mixtures, which lack a clean reference, with reference-available remixes of the same speakers; Track~2 additionally degrades the target video in five ways and adds 3-m far-field recordings. Sixteen and twelve teams were ranked on speaker-disjoint test data by rank averaging over waveform fidelity, predicted quality, transcription accuracy, and speaker similarity. On Track~1 remixes, the best system reaches 12.7~dB SI-SDR and 0.85 STOI, but natural recordings remain harder: even the lowest CER rises from 9.0\% to 14.7\%. The leading systems use video mainly for speaker attribution rather than signal reconstruction, and the top two lose under 0.5~dB SI-SDR on Track~2. UTMOS and DNSMOS rank systems differently from the other metrics, so no single metric captures target-speech recovery. We release the baselines, evaluator, and official results.

eess.AS↗

arXiv2026

Behavior-Aligned Action Tokenization for Robot Policy Learning

Autoregressive robot policies learn continuous control by predicting discrete action tokens from observations. Different tasks often share local motions, yet behavioral correspondence across demonstrations receives limited explicit supervision in existing tokenizers. Motions with different timing can therefore lack a shared representation despite following similar patterns. We propose Behavior-Aligned Action Tokenization (BAAT), which uses soft dynamic time warping (Soft-DTW) to select corresponding action chunks and aligns their quantized coordinates jointly with reconstruction. This objective encourages similar motions across tasks to occupy nearby quantized representations while retaining executable action detail. A history-conditioned diffusion decoder reconstructs continuous action chunks from these tokens, and a downstream autoregressive policy learns to predict them. We evaluate BAAT on selected tasks from three simulation benchmarks and two real robot tasks. BAAT achieves a mean simulation success rate of approximately 45.2%, exceeding OAT by approximately 7.2 percentage points. In the controlled LIBERO-All alignment ablation, policy success rises from 70.2% to 79.0% while trajectory replay success decreases. These results support behavioral correspondence as supervision for organizing shared motion structure in action tokenizers and improving downstream robot policy learning.

cs.RO↗

arXiv2026

AI Persona, Service Consumption, and User Intent Entropy: Field Experimental Evidence from an LLM Platform

Problem definition: Firms deploying large language model services must decide how their AI communicates, not just what it can do. We examine how a relational persona - warmer, more empathetic and more engaging than a non-relational persona - affects service consumption and the evolution of user objectives. Methodology/results: In a randomized field experiment with 9,586 newly registered users, we hold the underlying model and service capabilities constant. The relational persona increases interactions (sessions, +8.1%; duration, +10.6%; chat rounds, +24.2%; intent entropy, +5.8%) and outputs (files, +12.3%; distinct goals, +12.1%). Effects vary by entry intent. First-session effects are insignificant for Task Execution users. Socialization and Knowledge Exploration users show similar increases in chat rounds: Socialization increases intent entropy without more outputs, whereas Knowledge Exploration increases outputs without higher intent entropy. Modeling intent dynamics as a transition process, we find higher intent transition entropy for Socialization (+11.8%) but higher intent continuation probability for Knowledge Exploration (+12.8%), suggesting greater conversational breadth and persistence, respectively. In subsequent use, the relational persona increases aggregate chat rounds and outputs across all entry intents. Session count rises by 12.2% for Task Execution and 49.4% for Socialization, but not significantly for Knowledge Exploration. Effects on session count and intent entropy strengthen over time, whereas output effects remain stable. Managerial implications: AI persona is an operational design lever, not merely a presentation feature. Because more interactions do not uniformly generate more outputs, firms should evaluate interactions and outputs separately and consider matching persona to user intent, especially when added interactions consume costly computing resources.

cs.HC↗

arXiv2026

Uniform-in-time approximation and convergence of invariant measuresfor the damped stochastic Korteweg-de Vries equation

To quantitatively characterize the long-time dynamics of the periodic damped stochastic Korteweg--de Vries (sKdV) equation driven by additive noise, we investigate the uniform-in-time error estimates for a Lie--Trotter operator splitting approximation. This splitting combines the exact deterministic KdV flow with the exact Ornstein--Uhlenbeck solution map for linear damping and additive forcing. Long-time error analysis for the sKdV equation is highly nontrivial because of the interplay among the lack of dissipative smoothing, the loss of one spatial derivative arising from the Burgers-type nonlinearity, and the fluctuations of stochastic forcing. To overcome these difficulties, we develop new strategies based on exponential Lyapunov estimates, the continuous dependence estimate, and the local error decomposition. We establish strong and weak order $1$ convergence uniformly in time under sufficiently large damping, as well as strong order $1$ convergence on finite time intervals under a weaker damping condition. Under stronger assumptions, the splitting approximation admits a unique invariant probability measure on $H_0^2$, which approximates the invariant measure of the exact dynamics with order $1$. To the best of our knowledge, these constitute the first uniform-in-time strong error and quantitative invariant-measure approximation results for the periodic damped sKdV equation with additive noise. Our results provide an avenue for long-time simulation of the damped sKdV equation.

math.PR↗

arXiv2026

A Unified Uncertainty-Aware Back-End for Speaker Verification: Scoring, Normalization, and Calibration

Speaker verification back-ends commonly combine similarity scoring, score normalization, and calibration. However, speaker embeddings extracted from real-world utterances have trial-dependent reliability because of factors such as duration, noise, and channel variation. Existing uncertainty-aware methods primarily improve the speaker encoder or the initial similarity score, while the estimated uncertainty is typically not propagated through subsequent normalization and calibration. We represent each utterance by a speaker embedding, interpreted as a posterior mean, together with its covariance as an uncertainty estimate. We present a unified uncertainty-aware back-end comprising uncertainty-aware cosine scoring, uncertainty-aware AS-Norm (UAS-Norm), and uncertainty-aware Quality Measure Function calibration (UQMF). Covariance information is incorporated throughout this pipeline to adjust score scaling, cohort statistics, normalized-score combination, and calibration features. Experiments with ECAPA-TDNN and ResNet show consistent EER reductions and improved target--non-target separation across both architectures.

cs.SD↗

arXiv2026

Qubit-Based Benchmarking of InP HEMT LNAs: Readout Fidelity Versus Power Consumption

High-fidelity single-shot readout of superconducting qubits is essential for fault-tolerant quantum computation. Without a quantum-limited amplifier such as a Josephson parametric amplifier, the cryogenic high-electron-mobility-transistor (HEMT) low-noise amplifier (LNA) at the 4 K stage is the dominant noise source in the readout chain. HEMT LNAs are conventionally characterized by Y-factor measurements of the noise temperature $T_N$, independently of the quantum measurement they ultimately serve. Here we present a qubit-in-the-loop benchmarking method. The method uses the two-dimensional IQ histogram of single-shot readout as a direct figure of merit. We apply it to three state-of-the-art cryogenic InP HEMT LNAs with 53\%, 60\%, and 70\% channel indium content, together with a commercial reference amplifier. The 60\% and 70\% devices have similar $T_N$ in Y-factor measurements, both lower than the 53\% device. The 70\% device has the highest gain, but qubit readout SNR and assignment fidelity $\mathcal{F}_a$ instead peak at 60\%. Using this method, we map $\mathcal{F}_a$ against HEMT LNA dc power consumption for each device. $\mathcal{F}_a$ does not simply saturate with increasing power. For several devices it declines above a device-dependent power well below typical operating points. For the 60\% device, which has the highest overall $\mathcal{F}_a$, about 1~mW is needed to maintain $\mathcal{F}_a > 85\%$. Across all four devices, $\mathcal{F}_a > 80\%$ is reached with as little as about 0.3~mW, roughly an order of magnitude below typical operating points. These results establish a qubit-referenced benchmarking methodology and provide practical guidance for amplifier selection and bias in power-constrained multi-qubit readout systems.

quant-ph↗

arXiv2026

Efficient difference-in-differences estimation under partial interference with incremental propensity score policies

This paper develops efficient difference-in-differences (DID) estimation under partial interference with a cluster incremental propensity score (CIPS) policy. We define direct and spillover average treatment effects on the treated, establish their identification, and derive their efficient influence functions, from which we construct a cross-fitted estimator. Simulations evaluate its finite-sample performance. An application to China's New Rural Pension Scheme recovers the reduction in farmwork among pension recipients reported by the original county-level analysis, separately estimates a within-household spillover alongside the direct effect, and traces both effects across the policy parameter.

econ.EM↗

arXiv2026

Learning to Prefer Reliably: Error-Augmented Emotion Preference Optimization with Calibrated Fusion

Emotion preference learning uses pairwise comparisons between candidate descriptions to align multimodal large language models (MLLMs) with human judgments of open-ended emotion descriptions and to train reward models that capture human emotional preferences. However, conventional pairwise supervision is often sparse, typically providing only a single negative description for each positive description, and therefore offers limited coverage of the diverse ways in which an emotion description can be incorrect. In particular, models may be insufficiently exposed to semantically fluent but emotionally inconsistent descriptions. Beyond this data-level limitation, relying on a single MLLM judge introduces a distinct model-level concern: its judgments can be affected by model-specific biases when interpreting fine-grained or ambiguous multimodal emotional cues. To address these limitations, we propose Error-Augmented Preference Optimization (EAPO), a framework for improving the reliability of MLLM-based emotion preference judgment at both the data and model levels. First, we construct an error-augmented dataset by generating multiple controlled and emotion-aware negative descriptions from each preferred description. We then adapt multiple independent MLLM judges to this richer supervision and aggregate their preference margins using margin-calibrated soft fusion, which maps heterogeneous margins to a common scale before aggregation. Experiments on the MER2026-EmoPrefer Challenge dataset and our error-augmented dataset demonstrate that EAPO improves emotion preference prediction and enhances the robustness of MLLM judges when evaluating fluent descriptions that conflict with the video's multimodal emotional evidence. Our code is available at https://github.com/slash1028/EAPO-EmoPrefer.

cs.MM↗

arXiv2026

EvoHarmBench: Breaking Content Moderation with Iterative Human-Like Evasion

Existing evaluations of harmful content detection rely predominantly on static benchmarks, which struggle to reflect the interactive adversarial ecosystem of real-world content platforms where users continuously revise their expressions in response to moderation feedback. This mismatch creates a significant performance gap between offline benchmark scores and online deployment effectiveness. To the best of our knowledge, we present EvoHarmBench, the first dynamic adversarial evaluation framework for content moderation systems. The framework employs an iterative optimization loop that evolves evasion strategies at the semantic-cluster level, while simultaneously optimizing for evasion success and human readability. We systematically evaluate LLM-based defense models which are widely used in real world moderation systems. The evaluation covers 229 semantic sub-clusters across five violation categories, derived from 5,002 real-world adversarial samples collected from content platforms. Our experiments reveal substantial vulnerabilities even in leading commercial systems: after twelve optimization iterations, the attack success rate under readability constraints reaches 80.3% within SOTA LLM moderators. We will release the full benchmark data, evaluation framework, and code to encourage a shift from static benchmarking toward dynamic adversarial evaluation in content safety research.

cs.CL↗

arXiv2026

On the $\mathrm{In_{x}Ga_{1-x}As}$ channel noise in InP HEMTs from 4 K to 300 K

The InP high-electron-mobility transistor (HEMT) is indispensable for low-noise amplifiers (LNAs) in radio astronomy and quantum computing. The composition of the $\mathrm{In_{x}Ga_{1-x}As}$ channel in InP HEMT is known to influence the LNA noise performance. However, the various physical mechanisms responsible for noise generation are not fully characterized and understood. Here, we investigate the $\mathrm{In_{x}Ga_{1-x}As}$ channel noise from 4 K to 300 K for 100-nm gate-length InP HEMTs with channel indium content of 53\%, 60\% and 70\%. Channel noise was quantified by extracting the equivalent drain noise temperature $\mathit{T}_{d}$ using both on-wafer and LNA-based measurements, covering 40-300 K and 4-40 K, respectively. The 60\% indium channel InP HEMT exhibited the lowest channel noise across the full temperature range. The $\mathit{T}_{d}$ extracted from on-wafer characterization was found to obey a parabolic temperature dependence which predicted the $\mathit{T}_{d}$ at 4 K for all InP HEMTs in good agreement with LNA-based measurements. By expressing the channel noise as the sum of one thermal and one excess noise term, it was found that the former increased linearly with ambient temperature and dominated at 300 K. The channel noise at 4 K was determined by the excess noise term and exhibited a non-monotonic dependence on the channel indium content in the InP HEMT. The results suggest that the excess noise in the InP HEMT originates not only from temperature-independent shot noise but also from impact ionization and real-space transfer noise.

cond-mat.mes-hall↗