Search arXiv⌕ Search

arXiv subjects

Zheng Wang

Publications and source records attributed to Zheng Wang.

At least 19 recordsLinked to original sources
arXiv2026

Prefrontal scaling of reward prediction error readout gates reinforcement-derived adaptive behavior in primates

Reinforcement learning (RL) driven by reward prediction errors (RPEs) enables adaptive behavior across species. While midbrain-striatal encoding of RPEs is well-documented, how prefrontal circuits translate these signals into species-specific flexibility remains unknown. Integrating cross-species RL modeling, neuroimaging, and transcriptomics, we identified shared molecular and neuroanatomical infrastructure for RPE computation in humans and macaques. Despite this conserved foundation, humans demonstrated superior performance on deterministic and probabilistic reversal tasks. Single-trial decoding revealed equally robust dynamic RPE-to-choice prediction in both species; yet humans disproportionately recruited dorsal anterior cingulate (dACC) and dorsolateral prefrontal cortex (dlPFC). Interspecies alignment localized similarly predictive policy signals in macaque dACC/dlPFC that failed to guide behavior. Crucially, a within-species comparison in an independent human cohort demonstrated that only near-optimal performers exhibited strong prefrontal action decoding, contrasting with suboptimal individuals. These findings suggest that behavioral flexibility arises from prefrontal readout breadth - a hierarchical scaling principle that challenges efficient coding theories.

q-bio.NC↗
arXiv2026

XPress: Parallel Refinement for Diffusion Drafters in Speculative Decoding

Block-diffusion drafters like dFlash generate an entire block of draft tokens in a single forward pass, drastically reducing the overhead of multiple-token drafting in speculative decoding. The crucial final step of the single-pass discrete denoising process involves using the logit distribution at each position to sample conditionally independent tokens. The resulting draft is thus a set of per-position marginals, rather than a joint distribution: no draft token is guaranteed to depend on its predecessors. Such independently sampled marginals tend to produce sequences with tokens that are individually likely, but jointly improbable under the target model's distribution, which verifies each token conditionally. This can cause early rejection and limits acceptance length. To address this, we propose xPress as a means to restore the missing causality in diffusion drafters. xPress is a lightweight causal refiner that reconciles the whole diffusion block at once through parallel refinement, restoring and propagating causal dependencies across the draft without a token-by-token loop. On Qwen3-8B, across seven math, code, and chat benchmarks, xPress raises acceptance length by about 30% on average (up to +56%) and its end-to-end decoding throughput by about 1.3 on average (up to 1.7) compared to the original dFlash diffusion drafter.

cs.AI↗
arXiv2026

ACTIVE: A Dynamic Pareto-Frontier-Based Unified Index for Dual-Vector Query

Existing approximate nearest neighbor search (ANNS) methods generally assume that each entity has a single vector. In many real-world applications, however, an entity is associated with more than one vector, and queries may jointly consider two attributes. With different queries assigning different weights to each attribute, Pareto-frontier-based indices provide an effective solution for dual-vector retrieval by connecting to approximate nearest neighbors under all possible query preferences. However, existing Pareto-frontier-based approaches suffer from substantial inefficiencies in both index construction and dynamic updates, limiting their practical applicability. To address these issues, we propose ACTIVE, a dynamic Pareto-frontier-based DVQ index that supports efficient construction and updates while maintaining high retrieval performance under varying query preferences. It employs angle-constrained active-range computation and connectivity-loss-driven adaptive repair strategies to accelerate node insertion and deletion, respectively. Extensive experiments show that, at comparable (or slightly below) search accuracy, ACTIVE improves insertion throughput by $1.20$--$1.43\times$ over DEG, and deletion throughput by $1.64$--$2.06\times$ over neighbor-reconnection-based repair and by $22.64$--$39.51\times$ over reinsertion-based repair.

cs.DB↗
arXiv2026

Same Context, Different Agents: A Behavioral Study of Semantic Layers, Multi-Agent Orchestration and Memory Compaction in Data Agents

Answering questions over heterogeneous data means combining evidence from files, databases, documents and the web. Data agents automate this, but their success depends on context. Three context-engineering techniques (CETs) help by shaping it: semantic layers describe the data, orchestration splits work among agents, and compaction condenses interaction history. Existing studies validate each CET alone, leaving open how agent behavior shapes its effect. We present a comprehensive analysis of the three CETs, alone and combined, with seven large language models (LLMs) on 1,514 long-horizon tasks from five benchmarks, covering behavior, accuracy and cost. We find that data agents without CETs form three behavior groups: investigators, premature answerers that answer without inspecting data, and early stoppers that give up. Each CET delivers its known benefit, from more investigation to smaller prompts, but only the behavior group it matches turns that benefit into accuracy, such as premature answerers under orchestration. For investigators, the same CETs can hurt: the semantic layer stops them from looking, orchestration makes them overconfident, costing 5 percentage points, and compaction makes them repeat steps. Combining CETs yields smaller gains and losses than their sum. We derive guidance for choosing CETs by agent behavior, task demands and accuracy-cost trade-offs.

cs.CL↗
arXiv2026

Realtime-Bench: Can Frontier Models Solve Real-time Interactive Problems?

Real-time interaction requires agents to perceive evolving environments and act under temporal constraints, yet existing evaluations of real-time streaming models focus on conversational quality and streaming perception, leaving effective action under strict temporal constraints untested. We introduce Realtime-Bench, a suite of carefully curated dynamic tasks where delayed actions directly affect success. Under Wall Clock, where environments advance according to actual inference latency, the best success rate reaches only 44% and the best Pass^3 is only 36%, demonstrating that real-time interaction remains challenging. To separate serving latency from models' real-time streaming capability, we compare Wall Clock with Paused Clock, which removes inference delays as an idealized reference, and introduce Token Clock, which advances simulated time according to generated token counts, testing temporal awareness independently of physical serving speed. We find that temporal costs limit success in evolving environments and that higher reasoning effort has mixed effects under time pressure, reflecting a trade-off between decision quality and timely execution. Based on these results, we discuss future directions for improving real-time multimodal models, including adaptive reasoning with temporal awareness and continuous perception and action.

cs.AI↗
arXiv2026

ATLAS: A Multi-Agent Framework for Lifting Binary to Canonical Compiler IR

Lifting stripped, heavily optimized binaries to a canonical, pre-optimization LLVM intermediate representation (IR) is essential for downstream security analysis and legacy software migration when source code is unavailable. However, aggressive compiler optimizations rewrite control and data flow, remove structural metadata, and introduce type-erased low-level idioms, causing current lifters to produce malformed or semantically divergent IR. We present ATLAS, a multi-agent framework designed to reconstruct formally valid and canonical LLVM IR from optimized disassembly.ATLAS combines (i) Structure-aware, region-level retrieval, which utilizes pseudo-probe attribution to align optimized instructions with canonical IR exemplars, and (ii) a collaborative refinement loop that closes the gap between probabilistic generation and deterministic compiler constraints. Concretely, a Lift agent generates an initial IR draft conditioned on retrieved exemplars, an Advisor agent interprets LLVM verifier/compilation diagnostics and static checks to propose localized repair actions, and a Validation agent applies the edits, re-runs toolchain verification and (when available) runtime checks, and feeds the resulting diagnostics back to the Advisor. The agents iterate until the IR is verifier-clean and behaviorally consistent, or a budget limit is reached. On HumanEval-Decompile and MBPP, ATLAS outperforms seven baselines on lifting LLVM IR, achieving 91.8% re-executability and 89.4% CFG similarity across ARM, x86-64, and RISC-V binaries. Moreover, our evaluation demonstrates that ATLAS also outperforms the Binary to C to IR pipelines in pointer-flow precision and downstream analyzability. By recovering a well-formed, machine-centric representation, ATLAS enables practical cross-ISA recompilation and verifiable binary-only migration workflows.

cs.SE↗
arXiv2026

Exact Archimedean Flicker--Rallis Periods on Cohomological Minimal $K$-Types

For the symmetric pair $(\mathrm{GL}_n(\mathbb C),\mathrm{GL}_n(\mathbb R))$, we compute the exact Flicker--Rallis period of a distinguished vector in a cohomological minimal $U(n)$-type, normalized by the Gel'fand--Tsetlin pairing of Hara--Miyazaki--Namikawa. Our main tool is a weight-independent identity between the Whittaker period and a compact $O(n)$-average. We determine the proportionality constant using the archimedean newform formula of Humphries--Jo, and evaluate the compact functional on the normalized distinguished vector. This yields an explicit formula for the Flicker--Rallis period for every cohomological weight.

math.RT↗
arXiv2026

VibeTraining-Bench: Data-Trainer-Harness Co-Design for Iterative System Improvement

Recursive self-improvement (RSI) requires agents to diagnose system bottlenecks and coordinate improvements across training and inference. We refer to the agents' capability to jointly optimize training data, the trainer, and the inference harness as cross-component co-design. We introduce VibeTraining-Bench, a benchmark for evaluating this capability across tasks spanning terminal operation, software engineering, multi-turn tool use, and mathematical reasoning. Within each task, research agents start from an identical initial agent system under a fixed resource budget. Guided by validation feedback, they iteratively refine data, trainers, and inference harnesses, then select and submit a final agent system for evaluation on the test set. To prevent reward hacking, we enforce three safeguards: test data isolation, separate research and grading containers, and network restrictions for research agents. Our evaluation results show that frontier research agents substantially improve target agent systems. For example, on SWE-bench Verified, GPT-6-Astra raises the target agent system's performance from 32.9% to 53.7%, comparable to the human-developed instruct model's 55.4%. However, there is still room to improve their ability to co-design and iteratively refine data, trainer, and harness to achieve cumulative performance gains. We hope that VibeTraining-Bench will serve as a unified platform for advancing RSI, deepening our understanding of cross-component optimization for agent systems. Code and data are available at https://github.com/D2I-ai/VibeTraining-Bench.

cs.AI↗
arXiv2026

From Generic to Dedicated: A Novel Optimizer for Online Continual Learning

Online Continual Learning (OCL) requires models to learn from non-stationary data streams under a strict single-pass constraint, making them highly susceptible to catastrophic forgetting. While existing studies have explored various strategies from data, architecture, and optimization perspectives, the optimizer often directly uses generic approaches such as SGD and AdamW. In this work, we reveal a new link between optimization dynamics and OCL needs by recasting Polyak-Ruppert averaging as the engine of "plasticity" and Primal averaging as the anchor of "stability". This inspires our novel optimizer, SPIN (Stability-Plasticity INterpolation), which explicitly decouples plasticity and stability by interpolating between a fast-moving "plasticity" sequence and an adaptive "stability" sequence. We theoretically demonstrate that SPIN implicitly employs an inverse Hessian approximation, providing crucial Tikhonov regularization to damp noisy, single-sample Hessian estimates and mitigate forgetting. More importantly, SPIN can be seamlessly integrated into existing OCL methods by taking resulted gradients as input and replacing standard optimizers. Extensive experiments show that using SPIN instead of standard optimizers yields consistent and significant performance gains across multiple OCL approaches, including the challenging rehearsal-free setting and under strict memory constraints.

cs.LG↗
arXiv2026

Pair-Aware Optimization: Coordinating Exploration and Refinement for Jailbreaking LLM Agents

Large Language Model (LLM) agents are vulnerable to adversarial inputs that induce attacker-specified malicious behaviors, commonly referred to as jailbreaking. Existing optimization-based jailbreak attacks primarily employ first-order (FO) methods that leverage gradient information or zeroth-order (ZO) methods that rely on function-value feedback. However, their complementarity and optimization ordering remain underexplored. We systematically investigate four strategies: pure ZO, pure FO, ZO->FO, and FO->ZO, and develop a ZO-FO framework that explores adversarial prefixes through ZO optimization and refines their corresponding suffixes through FO optimization. To coordinate the two processes, we introduce Pair-aware optimization, which retains multiple prefix candidates, conditionally refines their suffixes and selects among the resulting prefix-suffix pairs. We further investigate agent jailbreaking from a more general stochastic-generation perspective, where the same adversarial input can induce diverse reasoning trajectories and actions that a single greedy output cannot capture. Accordingly, we adopt consistent stochastic decoding during both optimization and evaluation. Experiments across malicious-instruction and malicious-environment benchmarks demonstrate that hybrid optimization improves attack effectiveness over single-optimizer methods, while optimization order substantially influences performance.

cs.LG↗
arXiv2026

VELD-QA: A Large-Scale Dataset for Visual Evidence Grounding and Question Answering in Long Documents

Existing long-document QA benchmarks rarely combine large-scale human-authored questions with detailed visual evidence annotations, limiting diagnosis of evidence localization and interpretation failures. We present VELD-QA, a large-scale benchmark for Visual Evidence grounding and question answering in Long Documents. It contains 15,000 fully human-authored question-answer pairs across 5,859 real-world PDFs totaling 372,445 pages, balancing chart, table, and complex-layout tasks. Human annotations identify supporting pages and evidence elements and describe cited visual content, enabling separate retrieval and interpretation evaluation. We evaluate nine vision-language models under four evidence settings and seven retrieval-augmented and agentic systems. Our analysis identifies bottlenecks in evidence localization and interpretation. Retrieval-based methods substantially underperform the oracle setting with ground-truth supporting pages; full-document access does not close the gap. Localization alone is insufficient: oracle accuracy peaks at 81.7%, highlighting persistent visual interpretation and reasoning challenges. Enriching inputs with our visual-element annotations narrows the performance gap by 17-24%, demonstrating value beyond evaluation and motivating data enrichment for multimodal document QA. Our code is available at https://github.com/oracle-samples/veld-qa.

cs.CV↗
arXiv2026

GRAD: Graph-Structured Antibody Defense for LLM Agents

Memory poisoning and decomposition attacks distribute malicious evidence across the operations of large language model agents. Existing request-level defenses can miss such attacks because they assess individual inputs without connecting earlier information to later actions. Moreover, detectors must adapt as task contexts change, while updates to shared parameters can impair previously learned detection capabilities. To address these challenges, we propose GRAD, which combines graph-based evidence aggregation across operations with an extensible collection of detection experts. Its Semantic-Structural Understanding Module (SSUM) jointly encodes operation text, event attributes, and typed candidate relations in an execution graph to aggregate distributed attack evidence into a graph-level risk score. Its Continual Antibody Adaptation (CAA) adds experts on frozen encoder snapshots and routes inputs by observable graph signatures, acquiring and reusing detection capabilities as tasks arrive. Across Decomp, MCP, and MRT, GRAD achieves AUROCs of 97.19%, 77.59%, and 86.60%, respectively, outperforming Sequential Monitor and locally adapted Agent-Sentry and NeuroTaint. CAA improves final accuracy by 13.08 percentage points over Replay on the cross-dataset task stream. The gain remains 9.05 points with equal family/base-case weights within each task. Its compact drop-in gate performs local model inference in milliseconds without monitoring API calls.

cs.CR↗
arXiv2026

Cognition-Oriented Emotion Tracing from Causes to Consequences in Real-World Social Scenes

Affective computing has progressed from categorical emotion recognition to open-ended affective analysis with large multimodal models. Yet affective science describes emotion as an unfolding process shaped by appraisal, regulation, and social interpretation, which remains underexplored computationally. We propose TRACE, a cognition-oriented framework that formalizes an affective episode through three interrelated stages: Condition, Affect, and Effect, integrating observable cues with cognitive factors such as internal stance and regulation of emotional display. Based on this formulation, TRACE-Bench evaluates multimodal models in real-world social scenes through five tasks spanning grounded affect recognition, regulation decoding, cause reasoning, effect reasoning, and full-chain reconstruction, with 3,746 structured question-answer pairs over 646 videos. A matched human-model comparison reveals a substantial performance gap, while affect-specialized models also generally lag behind general-purpose MLLMs. Model outputs show recurring failures, including treating displayed behavior as genuine feeling and fabricating unsupported events during long-chain generation. We further propose TRACER, a cognition-grounded structured reasoning method that couples each inference with explicit premises from factual observations, cognitive appraisals, and established upstream conclusions, forming a traceable graph of intermediate and target conclusions. TRACER outperforms all evaluated model baselines on each of the five tasks. Project page: https://cogaffc.github.io/TRACE

cs.AI↗
arXiv2026

From Expected Harmfulness to Likelihood: A Probabilistic Reformulation of Jailbreaking LLM Agents

When the harmfulness of an LLM agent's output can be quantified, a natural jailbreaking objective is to maximize expected harmfulness over admissible input modifications. An alternative approach constructs or selects harmful target outputs and modifies the input to increase their likelihood. We establish a precise connection between these two approaches through a probabilistic reformulation. Specifically, we show that the gradient of the logarithm of expected harmfulness with respect to the input equals the expected input gradient of the model's log-likelihood under a harmfulness reweighted output distribution. This identity provides a unified interpretation of expected harmfulness and target likelihood optimization. Building on this connection, we propose OPUR, a sampling distribution designed to generate highly harmful target outputs and use the resulting samples to guide likelihood-based input optimization. Experiments demonstrate the effectiveness of the resulting method in jailbreaking LLM agents.

cs.AI↗
arXiv2026

Revisiting Cross-Reconstruction for Generalizable Deepfake Detection

Existing image forgery detectors often suffer from generalization to unseen manipulation methods due to the limited ability to capture transferable forensic cues. Recent cross-reconstruction based methods attempt to improve generalization through semantic-artifact disentanglement, but typically align heterogeneous artifacts across generators and exclude artifact representations during reconstruction, which may overlook the inherent diversity and visual cues of manipulation artifacts. In this work, we revisit cross-reconstruction and introduce an artifact-oriented disentanglement framework for robust image forgery detection. We argue that \textbf{artifact diversity}, i.e., the intrinsic variations of manipulation artifacts introduced by different generation processes, contains complementary forensic cues rather than undesirable domain variations. Instead of enforcing explicit artifact alignment, our framework preserves diverse artifact characteristics through semantically aligned cross-generator reconstruction. Furthermore, we incorporate artifact representations into the reconstruction process and introduce a masked frequency-aware reconstruction strategy to emphasize manipulation-related residuals while reducing semantic interference. This design enables the model to learn transferable forensic representations from diverse artifacts. Extensive experiments on multiple benchmark datasets demonstrate improvements under both cross-dataset and cross-generator evaluation settings. Further analysis and ablation studies validate the effectiveness of artifact diversity preservation and artifact-aware cross-reconstruction.

cs.CV↗
arXiv2026

GeoLatent: Geometry-Guided Latent Structuring with Routed Optimization for 3D Reasoning

Despite progress in vision-language models, 3D spatial reasoning from 2D images remains challenging. Text-based methods describe intermediate geometry with discrete tokens, limiting fidelity for continuous spatial relations. Continuous latents offer richer representations, but a single latent type does not explicitly separate the cues needed across spatial tasks. Decomposed spatial latents address this by representing position, direction, and global geometry separately under geometric supervision. Yet the geometry representation can still collapse toward one dominant direction, and unrestricted attention can leave the latents underused during answer learning. We introduce GeoLatent, combining Common--Residual Geometry Alignment (CR-GEO) with routed optimization to structure the geometry states while promoting latent-mediated answer learning. CR-GEO separates shared from residual teacher geometry; routed optimization jointly trains geometry and language, temporarily directs visual answer learning through the latents, and restores full attention with geometry supervision. In controlled comparisons, CR-GEO raises geometry effective rank from 1.00 to 3.87, while blocking latent readout at the bottleneck lowers direction accuracy from 89.1% to 25.8% on 128 fixed questions. After recovery, the differentiated geometry representation and latent-mediated visual route remain available alongside direct image access. GeoLatent achieves 73.0% on SPAR-Bench and 72.1% on SPBench, outperforming previously reported methods on both.

cs.CV↗
arXiv2026

Large-gap quantum anomalous Hall insulator in two-dimensional pentagonal TiN$_8$ monolayer

Discovering the quantum anomalous Hall (QAH) effect in two-dimensional magnets is a central goal of topological condensed matter physics, yet candidate materials combining a sizable gap with a well-defined Chern number remain scarce. We report a new material platform that realizes emergent magnetism and nontrivial band topology within the largely overlooked family of two-dimensional MN$_8$ monolayers. Through first-principles calculations, we reveal that these systems exhibit out-of-plane magnetic ground states and nontrivial topological features driven by the localized $d$-orbitals of the embedded transition metals. Remarkably, we identify TiN$_8$ as a quantum anomalous Hall insulator characterized by a Chern number of $C=-1$ with a large topological gap of 223 meV. As a companion system, MoN$_8$ is predicted to be a rare high-Chern-number QAH insulator, with a Chern number of $C=2$, although its much smaller gap of 12 meV renders it more fragile. These findings establish the penta-MN$_8$ family as a versatile and promising platform to realize exotic topological quantum states. This work not only broadens the material landscape for magnetic topological insulators but also lays a solid theoretical foundation for the development of next-generation spintronic and quantum computing devices.

cond-mat.mtrl-sci↗
arXiv2026

Exploiting Vulnerabilities: Universal Adversarial Attacks on Vision-Language-Action Models in Robotics

Recently, Vision-Language-Action (VLA) models have revolutionized robotic manipulation by seamlessly integrating visual perception, language understanding, and action generation in an end-to-end learning framework. However, since these models are designed to interact directly with the physical world and humans, their security is critical, and even small vulnerabilities can lead to catastrophic failures. In this work, we propose the Universal Adversarial Object, a sphere with optimized surface texture that significantly degrades task success rates when placed within the robot's field of view. Specifically, our approach introduces a multi-level attack framework that jointly disrupts trajectory planning, task execution, and action control. We validate our method in both simulated and real-world robotic settings. Experimental results demonstrate that the adversarial object reduces the average task success rates by 31.2%-39.9% for two representative VLA models (Pi0 and RDT), with success rates dropping to near zero in complex scenarios. Index Terms--Vision-Language-Action models, adversarial attack, robotic security, universal adversarial object

cs.RO↗