arXiv subjects

Lu Wang

Publications and source records attributed to Lu Wang.

At least 37 records · Page 2Linked to original sources

arXiv2026

SchemeArena: Factorized Stress Testing of Scheming in LLM Agents

We study scheming in LLM agents, in which agents covertly pursue misaligned goals. Our focus is to understand how scheming arises from the interaction of key factors, such as instrumental goals, environmental affordances, oversight conditions, and perceived consequences. Prior work examines only a small number of scenarios, limiting the ability to isolate how these conditions shape an agent's propensity or capability to scheme. This limited scale and task diversity also restrict coverage of realistic deployment settings and the range of scheming strategies that can be observed. To this end, we introduce SCHEMEARENA, a 400-scenario benchmark for scalable scheming stress testing, constructed through a factorized scenario synthesis framework spanning diverse safety-relevant tool domains, instrumental goals, oversight conditions, and pressure mechanisms. To enable scalable and reliable monitoring, we further propose SCOUT, a scheming monitor that grounds multi-criteria judgments in evidence drawn from agents' reasoning and actions. Across controlled stress tests on five LLM agents, we find that explicit instrumental goals are the strongest driver of scheming propensity. Strategic hints play a distinct role by helping agents translate scheming reasoning into concrete covert behavior. Oversight has mixed effects: in several closed models, action-only monitoring increases scheming, suggesting that partial oversight can act as an optimization constraint rather than a deterrent. CoT is a useful but incomplete monitoring signal: it can reveal latent scheming before execution, yet action-only scheming shows that covert behavior may occur without explicit reasoning evidence. We release the benchmark, code, and monitor at: https://github.com/launchnlp/SchemeArena.

cs.AI↗

arXiv2026

Beyond State Consistency: Behavior Consistency in Text-Based World Models

World models have been emerging as critical components for assessing the consequences of actions generated by interactive agents in online planning and offline evaluation. In text-based environments, world models are typically evaluated and trained with single-step metrics such as Exact Match, aiming to improve the similarity between predicted and real-world states, but such metrics have been shown to be insufficient for capturing actual agent behavior. To address this issue, we introduce a new behavior-aligned training paradigm aimed at improving the functional consistency between the world model and the real environment. This paradigm focuses on optimizing a tractable step-level metric named Behavior Consistency Reward (BehR), which measures how much the likelihood of a logged next action changes between the real state and the world-model-predicted state under a frozen Reference Agent. Experiments on WebShop and TextWorld show that BehR-based training improves long-term alignment in several settings, with the clearest gains in WebShop and less movement in near-ceiling regimes, while preserving or improving single-step prediction quality in three of four settings. World models trained with BehR also achieve lower false positives in offline surrogate evaluation and show modest but encouraging gains in inference-time lookahead planning.

cs.LG↗

arXiv2026

Lightweight Soft X-ray Imager (LSXI) with glass-based coded mask

The coded mask technique has been widely used in X-ray and Gamma-ray imagers, especially in space astronomy. However, the traditional design of coded mask imagers usually has problems with large size and heavy weight. Here, we propose a novel design for a coded mask imager made of glass based on microchannel plate (MCP) technology, making it lightweight (within 1 kg), compact, and self-supporting, which is very suitable for space exploration satellites. The design is initially demonstrated by reconstructing encoded patterns from measurements at the X-ray beamline. Monte Carlo simulations of the module integrated into a 6U CubeSat are conducted to assess its in-orbit detector performance, including sensitivity and source localization.

astro-ph.IM↗

arXiv2026

Quantum Dial for High-Harmonic Generation

High-harmonic generation (HHG) is a highly nonlinear optical process that typically requires an intense laser to trigger emissions at integer multiples of the driving field frequency. However, the strong fields required for conventional HHG inevitably perturb the system, limiting its use as a nondestructive spectroscopic probe. Recent advances in bright squeezed vacuum (BSV) sources have created opportunities to drive HHG with quantum fields alone. In this work, we demonstrate a regime in which the light-matter interactions can be controlled and tuned using a weak classical field, whose pulse energy is two orders of magnitude lower than that in standard HHG-perturbed by an even weaker quantum field such as BSV. This approach opens new avenues for nonlinear spectroscopy of materials while substantially suppressing strong laser-induced damage, distortions, and heating. We show that a BSV pulse containing less than 5% of the classical driving energy can act as an 'optical dial', allowing tuning of the nonlinear emission spectrum, emission angular dependence, and ionization.

physics.optics↗

arXiv2026

Co-Evolving Actor-Conditioned Critics for Non-Verifiable Generation

Natural-language critiques provide supervision beyond scalar rewards for non-verifiable generation, which lacks deterministic verifiers. In critique-guided refinement, a critic gives feedback on an initial response and an actor revises it. However, final revision quality does not reveal whether the critique was actually useful: a capable actor may improve without following the feedback, while valid feedback may fail if the actor cannot execute it. We frame critique as actor-conditioned revision guidance, where usefulness depends on whether the feedback helps the target actor address the intended weakness. We introduce TAIScore (Targeted Actionable Improvement Score), a reward that evaluates the instruction, initial response, critique, and revision together, assessing whether the critique targets a real weakness, whether the actor follows it, and whether the intended aspect improves. We use this reward to train an actor-tailored critic with GRPO, and use critique-guided refinements to construct DPO preference pairs for the actor, forming a co-evolving critic-actor loop where the critic adapts to the actor's changing capability. Experiments show that an 8B critic trained with TAIScore outperforms both a zero-shot 120B critic and critics trained with outcome-only or critique-only reward signals. Co-evolving the critic and actor further improves performance, suggesting that effective critique supervision should adapt as the actor changes.

cs.CL↗

arXiv2026

MoPLEx: Estimating Plackett-Luce Mixture Models for Multi-Objective Alignment

We study learning a mixture of $k$ Plackett-Luce models from multi-way ranking responses from annotators that may represent heterogeneous underlying preferences. This problem has many applications in AI alignment and preference optimization. Prior work has studied mixtures of Bradley-Terry models from pairwise comparisons. However, estimating a mixture of multi-way ranking models can become theoretically unidentifiable when $k$ exceeds $m/2$, where $m$ is the ranking length. We design an efficient algorithm to address this issue by first augmenting the rankings to a larger size (e.g., generating comparisons from a base model), followed by a gradient-based estimation to reduce inference cost (in the input embedding space). With this procedure in mind, we then fit a mixture of Plackett-Luce (PL) models via an expectation-maximization-style iteration, or MoPLEx in short. We conduct extensive experiments to verify this algorithm. First, we find that the gradient-based approximation estimates true probabilities with less than 5% error on models with up to 34 billion parameters. Second, MoPLEx improves clustering and ranking accuracy by an average of 43.7% and 15.2% over baselines using a single PL model or a mixture of Bradley-Terry models, on UltraFeedback and PERSONA datasets. These results demonstrate the effectiveness of MoPLEx for tackling multi-way rankings following heterogeneous preferences through measuring alignment via gradients.

cs.LG↗

arXiv2026

Can LLM Agents Discover? Evaluating Creativity on ML Engineering Tasks

Recent AI systems promise autonomous scientific discovery, claiming to discover algorithms and produce research papers, yet understanding whether they exhibit creativity, the capacity to produce solutions that are both novel and useful, remains an open question. We present a framework for evaluating multi-turn LLM research agents' creativity using ML engineering tasks as a testbed, through three dimensions: P-Creativity (psychological novelty: novel relative to the agent's own prior solutions within a run), H-Creativity (historical novelty: novel relative to the corpus of human solutions), and Usefulness (task performance). Evaluating two agent frameworks, AIDE and AIRA-Dojo, on 10 Kaggle-style machine learning tasks from MLE-Bench, we develop an LLM-as-a-Judge pipeline and verify its strong correlation with human creativity judgments, providing a reliable automated metric for P-Creativity evaluation at scale. Applying this pipeline to agent trajectories, we find: (1) all agents exhibit declining P-Creativity as they transition from exploration to exploitation; (2) LLMs exhibit greater H-Creativity than medal-winning humans, yet achieve lower performance. Our findings reveal that current agents can explore novel regions of the solution space but lack the capacity to convert this novelty into improved task performance.

cs.AI↗

arXiv2026

CASPER in the Machine: Insights into Character Variety in LLM-Generated Stories

As LLM-generated text is increasingly used, especially in fictional domains, we explore how much LLM-generated stories differ from human-written stories. In this work, we focus on characters. We borrow definitions from narratology to analyze eight intricate dimensions of character, such as stylization and wholeness. These dimensions consider more than just basic characteristics. They assess how characters are portrayed within their stories. After automatically inferring categories of characters within both LLM and human-written stories, we compare and contrast these two sets of stories. We consider the following overarching questions: (1) Do LLMs and human-written stories have similar characters? and (2) Do LLMs generate stories with a variety of characters? Our analysis includes research questions that focus on stories generated by popular LLMs and recently published human-written stories. We describe a number of interesting similarities, differences and key takeaways.

cs.CL↗

arXiv2026

IndexTTS 2.5 Technical Report

In prior work, we introduced IndexTTS 2, a zero-shot neural text-to-speech foundation model comprising two core components: a transformer-based Text-to-Semantic (T2S) module and a non-autoregressive Semantic-to-Mel (S2M) module, which together enable faithful emotion replication and establish the first autoregressive duration-controllable generative paradigm. Building upon this, we present IndexTTS 2.5, which significantly enhances multilingual coverage, inference speed, and overall synthesis quality through four key improvements: 1) Semantic Codec Compression: we reduce the semantic codec frame rate from 50 Hz to 25 Hz, halving sequence length and substantially lowering both training and inference costs; 2) Architectural Upgrade: we replace the U-DiT-based backbone of the S2M module with a more efficient Zipformer-based modeling architecture, achieving notable parameter reduction and faster mel-spectrogram generation; 3) Multilingual Extension: We propose three explicit cross-lingual modeling strategies, boundary-aware alignment, token-level concatenation, and instruction-guided generation, establishing practical design principles for zero-shot multilingual emotional TTS that supports Chinese, English, Japanese, and Spanish, and enables robust emotion transfer even without target-language emotional training data; 4) Reinforcement Learning Optimization: we apply GRPO in post-training of the T2S module, improving pronunciation accuracy and natrualness. Experiments show that IndexTTS 2.5 not only supports broader language coverage but also replicates emotional prosody in unseen languages under the same zero-shot setting. IndexTTS 2.5 achieves a 2.28 times improvement in RTF while maintaining comparable WER and speaker similarity to IndexTTS 2.

cs.SD↗

arXiv2026

StanceFlip: A Comprehensive Multi-Dimensional Benchmark for Multimodal Conversational Stance Flipping Forecasting

Conversational stance detection has shifted from static text analysis to dynamic multimodal modeling. However, existing benchmarks exhibit three key limitations: failure to capture the dynamic evolution of beliefs, particularly during stance reversals; difficulty in disentangling affective states from logical reasoning; and neglect of the critical role of multimodal cues in resolving pragmatic ambiguities such as sarcasm. To address these limitations, we propose StanceFlip, a benchmark designed for multimodal conversational stance flipping forecasting over multi-turn dialogues across five modalities and multi-scenarios, which includes two novel subtasks: 1) Multimodal Stance Sextuple Extraction, extracting holder, target, emotion, sentiment, stance, and rationale as static state snapshots of dialogue to capture fine-grained cognitive structures. 2) Dynamic Stance Flip Attribution, tracking stance reversals across the conversation and identifying their underlying triggers. Alongside the dataset, we propose a dedicated framework, named ConStaFF, for Multimodal Conversational Stance Flipping Forecasting (MCSFF). Built upon a large language model, ConStaFF performs end-to-end stance reasoning, with a Thought-of-Stance (ToS) reasoning framework and a self-reflective verification mechanism integrated for structured stance modeling and faithful flip attribution. Specifically, ToS decomposes the reasoning process into specialized cognitive personas to formulate target propositions, resolve cross-modal conflicts, and infer historical stance trajectories. Extensive experiments show that our approach achieves state-of-the-art performance on both sextuple extraction and flip-trigger attribution, outperforming strong multimodal large language model baselines by substantial margins.

cs.CL↗

arXiv2026

EvoSpec: Evolving Speculative Decoding via Real-Time Vocabulary and Parameter Adaptation

Speculative decoding accelerates Large Language Model inference through draft-then-verify generation, yet lightweight draft models face coupled efficiency and quality limitations: large-vocabulary output projection is costly, while limited draft capacity and static parameters reduce acceptance under specialized or shifting inputs. Vocabulary pruning lowers projection cost, but static variants miss locally important long-tail tokens, while dynamic variants remain sensitive to preset selection policies and budgets. Moreover, limited draft capacity can leave the draft distribution misaligned even when the target token is covered. Online alignment improves draft quality, but full-parameter updates introduce substantial memory and latency overhead. We introduce EvoSpec, which jointly adapts the active vocabulary and lightweight draft parameters from verification feedback. EvoSpec asynchronously retrieves semantic and statistical token neighbors and performs curriculum-weighted online LoRA alignment while preserving exact target-model verification. On Qwen3-8B/EAGLE-2, EvoSpec reaches a $2.18\times$ speedup over vanilla decoding and a $1.20\times$ gain over EAGLE-2, while improving specialized-domain coverage and using $27\%$ less auxiliary GPU adaptation memory than full-parameter online adaptation.

cs.CL↗

arXiv2026

Beyond Fail-to-Pass: Iterative Hardening of Co-Generated Bug Reproduction Tests and Fixes

Large language models (LLMs) have made automated program repair (APR) increasingly practical for real-world bugs, but repairing directly from bug reports remains underconstrained. Bug reproduction tests (BRTs) help close this gap by turning a bug report into an executable, bug-specific signal that can guide repair and validate candidate patches. Existing work has therefore studied BRT generation as a core subproblem in APR and mainly evaluates a generated BRT using the fail-to-pass (F->P) criterion, which requires the test to fail on the buggy code but pass on the golden fix. We show that F->P alone is insufficient when the goal of a BRT is to improve downstream repair. In particular, some F->P BRTs are lax, reproducing the observed symptom yet still admitting plausible-but-incorrect patches. We formalize this missing quality dimension by separating F->P BRTs into rigorous and lax ones, and show empirically that only the former consistently improve repair success. We further find that co-generation introduces test--fix error coupling, where the in-trajectory fail-to-pass (F->P) check can pass even when both the generated patch and generated test are wrong. Based on these findings, we propose CoHarden, a co-generation framework that uses the Lax signal as an in-loop convergence criterion. CoHarden first generates a test before any fix, then iteratively hardens the test and fix against surviving mutation patches until the generated test no longer admits Lax regressions. Experiments show that CoHarden reaches 69.4% Resolved and 78.9% F->P on SWE-bench Verified, outperforming the strongest fix-only and cogeneration baselines by +9.6 and +7.9 percentage points in Resolved, respectively, with consistent gains across LLM backbones and benchmarks.

cs.SE↗

arXiv2026

Implicit Actor Critic Coupling via a Supervised Learning Framework for RLVR

Recent advances in Reinforcement Learning with Verifiable Rewards (RLVR) have empowered large language models (LLMs) to tackle challenging reasoning tasks such as mathematics and programming, however existing RLVR methods often suffer from sparse reward signals and unstable policy gradient updates inherent to RL-based approaches. To address the challenges, we propose PACS, a novel RLVR framework that achieves imPlicit Actor Critic coupling via a Supervised learning framework. By treating the outcome reward as a predictable label, we reformulate the RLVR problem into a supervised learning task over a score function parameterized by the policy model and optimized using cross-entropy loss. A detailed gradient analysis shows that this supervised formulation inherently recovers the classical policy gradient update while providing more stable and efficient training. Extensive experiments demonstrate that PACS significantly outperforms strong open-source models and RLVR baselines, yielding substantial average gains of +8.26% (4B) and +9.57% (8B) over base models offering a promising avenue for LLMs post-training with verifiable rewards. Our code and data are available as open source at https://github.com/ritzz-ai/PACS.

cs.CL↗

arXiv2026

MET: Theory-Grounded and Culture-Aware Multilingual Moral Reasoning

Language models are increasingly used for moral decision-making across diverse linguistic and cultural contexts, yet existing work overlooks multilinguality on three aspects: 1) multilingual evaluation benchmarks use direct translation, failing to adapt culture-specific items; 2) inference-time methods for moral reasoning rely on static, English-centric scaffolds and lack grounding in moral theory; 3) training methods for moral decision-making typically require expensive supervision from stronger models or human annotators. We address these gaps with three contributions. First, we introduce MCLASH, a multilingual moral decision-making benchmark to capture culturally situated moral intuitions and social norms across languages. Second, we propose MET (Multilingual Ethics with Theory-grounded reasoning), a two-step prompting method built on expert-curated, theory-based grounds drawn from psychology and philosophy: the model first selects situation- and culture-specific grounds, then reasons over them in the native language of the user. Third, we introduce MET-D (MET-Distillation), which enhances the second step through a self-distillation training stage that requires no external supervision. MET-D improves macro-F1 over the base model on all three models of different sizes and families (Qwen3-4B, Qwen3-8B, Gemma3-4B), by an average of 3.71 points on MCLASH and 4.23 on MMoralExceptQA, with a peak MCLASH gain of 12.94 points for Malay on Qwen3-8B. We further reveal that MET-D increases native-language reasoning by 62.13 points on average, and that beneficial grounds differ systematically across cultures. Together, these contributions open the path for culture-aligned, theory-grounded multilingual moral reasoning.

cs.CL↗

arXiv2026

Active rejection enables reliable generalization of universal machine-learning interatomic potentials

Universal machine learning interatomic potentials (uMLIPs) bridge quantum-mechanical accuracy and large-scale molecular dynamics, but the cost of high-accuracy calculations such as r$^2$SCAN limits training to datasets that remain small relative to the open materials space. Strong average benchmark performance also does not guarantee reliable energy--force predictions for every structure. We propose Adaptive Multi-Teacher Routing (ATR), which reformulates high-fidelity data construction as a structure-wise decision problem under uncertainty. Using a small set of real r$^2$SCAN labels, ATR calibrates multiple pretrained uMLIP teachers and combines structural descriptors, teacher identity, and inter-teacher disagreement to estimate the reliability of each structure--teacher pair. It selects high-confidence predictions for pseudo-label generation and rejects structures for which no teacher is sufficiently reliable. With real r$^2$SCAN labels for only 0.2\% of candidate structures, ATR distils 2.89 million traceable r$^2$SCAN-level pseudo-labels for pretraining. On held-out r$^2$SCAN structures and the MP-r$^2$SCAN benchmark, a lightweight CHGNet trained on the ATR-generated dataset consistently outperforms the baseline and non-routed controls. Finite-temperature molecular dynamics further shows that ATR improves dynamical robustness across multiple material systems, maintaining stable trajectories where baseline simulations undergo catastrophic structural collapse. These results establish active rejection as an effective mechanism for converting multiple pretrained uMLIPs into a scalable and reliable data-construction system for high-fidelity uMLIPs.

cs.LG↗

arXiv2026

LiveOIBench: Can Large Language Models Outperform Human Contestants in Informatics Olympiads?

Competitive programming problems are increasingly used to evaluate the coding capabilities of large language models (LLMs) due to their complexity and ease of verification. Yet, current coding benchmarks face limitations such as a lack of exceptionally challenging problems, insufficient test case coverage, and reliance on online platform APIs that limit accessibility. To address these issues, we introduce LiveOIBench, a large-scale competitive programming benchmark featuring 403 expert-curated problems, averaging 60 official test cases each, drawn from 72 contests across 14 Informatics Olympiads held between 2023 and 2025. LiveOIBench has four key features: (1) expert-designed tasks with detailed subtask rubrics and extensive test cases; (2) direct comparison to elite human contestants; (3) continuous updates to reduce contamination risk; and (4) a fully offline, reproducible evaluation system. Benchmarking 34 popular general-purpose and reasoning LLMs, we find that GPT-5 achieves an 81.76th percentile, still falling short of top human contestants, while among the open-weight models, GPT-OSS-120B reaches only the 60th percentile. Reasoning-trace analyses indicate that robust reasoning models prioritize precise problem analysis over excessive exploration. Finally, analyses across release dates, task familiarity, and code similarity find minimal evidence of data contamination in our benchmark. Our leaderboard, code, and data are available at: https://liveoibench.github.io/.

cs.AI↗

arXiv2026

Benchmarking the Robustness of Autonomous Driving to Environmental Illusions: A Lane Perception Perspective

Environmental illusions (eg., shadows, reflections, and tire marks) are naturally existing yet overlooked phenomena in real-world driving environments. They can disturb visual perception, leading to misinterpretation of the scene and posing serious safety risks to autonomous driving (AD) systems. However, existing researches largely overlook these phenomena, leaving a critical gap. To address this issue, we study AD robustness through the lane perception perspective, a fundamental task supporting core functions like cruise control and lane centering. We focus on two representative models: conventional lane detection (LD) and vision-language model-based systems (ADVLMs). In this work, we introduce the first benchmark, LanEvil++, for evaluating the robustness of lane perception under environmental illusions. LanEvil++ encompasses 14 types of illusions and leverages the CARLA simulator to generate 94 high-fidelity, fully controllable 3D scenes, yielding a dataset of 90,292 annotated images, 1,596 video clips, and 41,855 visual question answering pairs. Extensive evaluations demonstrate that environmental illusions substantially degrade the performance of state-of-the-art LD methods. On average, LD models experience a 5.27% drop in Accuracy and a 10.49% decline in F1-score, while ADVLMs show a 2.03% reduction in GPT-score and a 0.75% drop in Language-score. Among all illusions, shadows emerge as the most disruptive factor, reducing accuracy by up to 7.20%. Furthermore, closed-loop simulations reveal that these illusions can lead to incorrect driving decisions. Complementary real-world case studies highlight safety-critical failures in actual traffic scenes. To enhance robustness, we propose the Multimodal Illusion Defense Approach (MIDA). MIDA achieves substantial gains under challenging conditions, boosting robustness by 4.23% on LD models and 3.82% on ADVLMs.

cs.CV↗

arXiv2026

FED-FSTQ: Fisher-Guided Token Quantization for Communication-Efficient Federated Fine-Tuning of LLMs on Edge Devices

Federated fine-tuning provides a practical route to adapt large language models (LLMs) on edge devices without centralizing private data. However, in mobile deployments, the training wall-clock is often dominated by straggler-limited uplink communication under heterogeneous bandwidth, intermittent participation, and non-IID client data. Although parameter-efficient fine-tuning (PEFT) methods such as LoRA and QLoRA reduce local memory and trainable parameters, repeated transmission of adapter updates remains a major bottleneck. We propose Fed-FSTQ, a semantic-sensitivity-aware communication-control primitive for communication-efficient federated LLM fine-tuning. Fed-FSTQ uses a lightweight token-level Fisher proxy to estimate semantic sensitivity, couples token-guided sparsification with mixed-precision adapter-update quantization, and allocates higher communication fidelity to semantically load-bearing evidence while suppressing redundant transmission. The method is drop-in compatible with standard federated PEFT pipelines and requires no change to the server aggregation rule. Experiments on multilingual QA and medical QA under non-IID partitions show that Fed-FSTQ reduces cumulative uplink traffic required to reach a fixed quality threshold by 46-fold relative to a Fed-LoRA baseline and improves straggler-limited wall-clock time-to-accuracy by 52%. Under the corrected Controlled LTE-20Mbps accounting, Fed-FSTQ reduces per-round time from 414.60s to 67.29s and reduces per-round energy from 839.20J to 146.28J, yielding a 6.16-fold speedup. On NVIDIA Jetson-class edge devices, Fisher-guided token reduction also yields up to a 1.55-fold inference speedup, demonstrating deployability under tight resource constraints.

cs.LG↗