arXiv subjects

Yufeng Wang

Publications and source records attributed to Yufeng Wang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Counterexamples to the Henning--Yeo Conjecture: Unbounded Fixed-Degree Gaps and Sharp First-Order Asymptotics

Henning and Yeo conjectured an upper bound on the identifying vertex cover number of a graph in terms of its order, size, and maximum degree. A two-parameter family $H_{t,r}$ of connected diameter-two graphs disproves the bound for every maximum degree at least four; after denominators are cleared, its margin is exactly $-(t-1)(r-1)$. The complement relation $τ_D=n-ρ$ exposes the mechanism: diameter-two fibres admit at most one packing vertex, while degree deficit accumulates under tree gluing with controlled port loads. Writing $A_Δ$ for the supremal additive gap at maximum degree exactly $Δ$, an exact transfer formula gives $A_Δ=+\infty$ for every $Δ\ge 4$, using Petersen fibres in degrees four and five and the original $H_{t,r}$ blocks in higher degrees. If $c_Δ$ denotes the corresponding supremal gap per vertex, rooted rook-graph fibres match a universal square-graph packing bound to first order. Consequently, $c_Δ\sim 1/Δ$, equivalently $Δc_Δ\to 1$ as $Δ\to\infty$.

math.CO↗

arXiv2026

Preference-Driven Online Adaptation for Personalized Interaction Initiation in Proactive AI Assistants

AI assistants are typically reactive, relying on users to initiate interactions. Proactive assistants go beyond this paradigm by autonomously initiating interactions based on users' activity contexts. However, appropriate interaction timing is user-specific and difficult to determine in advance, while online feedback offers valuable signals for personalization. Direct feedback-driven adaptation is therefore appealing, but remains challenging due to sparse interaction-worthy moments scattered across fine-grained user states. To address the issues, we propose Evidence-driven Online Preference Adaptation (EOPA), which grounds a user's interaction-timing preferences in measurable contextual evidence through two evidence carriers: temporal preference anchors and evidence-bearing activity prototypes. At each polling step, EOPA derives temporal and activity evidence from the carriers through user-prior-smoothed evidence estimation and uncertainty-guided evidence scaling, and adaptively fuses the evidence for interaction-or-silence decisions. When interaction is selected, an LLM uses high-quality historical responses as demonstrations to generate a context-aware response that better reflects user preferences. EOPA updates its evidence carriers and decision parameters from received online feedback without LLM-based reasoning or retraining. Extensive experiments on a ProPerSim-based benchmark show that EOPA improves the interaction-timing F1 score by 19.80 points over the strongest baseline in our experiments, substantially reduces inference latency for both silence and interaction steps, and lowers the average daily adaptation time from 11.41 to 0.39 seconds.

cs.HC↗

arXiv2026

BulkPR-Bench: Benchmarking Queue-Level Governance of Interacting Pull Requests

Coding-agent benchmarks increasingly cover long-horizon, end-to-end, and interactive development, but typically retain one requested outcome or a fixed change sequence. Sequential policies can process a pull-request (PR) queue one candidate at a time, but when queued PRs interact, maximizing safe delivery can require jointly deciding which changes to merge and in what order. We introduce BulkPR-Bench, an executable benchmark in which an agent must recover consequential PR relations and return a large safe subset in executable order under a rolling-release protocol. The suite contains 581 newly authored candidate PRs on frozen snapshots of 18 real repositories. Registered state-by-state repository execution, including hidden safety checks, validates the gold relation graph; an exact oracle then computes the largest safe subset. Our primary metric, Relational Delivery Score (RDS), scores safe delivery and correct rejection over relation groups from the realized merge trace; Global Safety-Gated Yield (Global-SGY) separately measures strict delivery of the realized whole-queue plan. Under the buffered primary protocol with batch size $K=32$, the three highest RDS estimates among the six models are 66.6%, 62.0%, and 57.9%, compared with 53.1% for the strongest sequential baseline. Only 8 of 324 model runs complete a queue exactly. Critical-relation recall ranges from 35.2% to 57.7%, and diagnostic runs supplied with the gold relations show substantial remaining headroom. Gains on relation groups therefore do not yet translate into dependable whole-queue governance.

cs.SE↗

arXiv2026

FirstResearch: Auditable Question Formation for LLM Scientific Discovery Agents

LLM systems for scientific discovery increasingly assist with ideation, literature synthesis, experiment planning, and report generation, but the first research question they propose can remain difficult to audit: it may sound plausible without exposing the mechanism, falsifier, or assumption that a scientist should inspect. We introduce FirstResearch, a first-principles research-question formation framework for scientific LLM agents whose core artifact is a structured Research Question Certificate. The certificate records primitive definitions, assumptions, a mechanism model, a tension or contradiction, a falsifiable hypothesis, a minimal decisive test, and a failure update rule, making the proposed question inspectable before downstream execution. On ten LLM-agent research topics, FirstResearch outperforms controlled prompt-level baselines inspired by AI co-scientist, Agent Laboratory, and AI Scientist-v2 under a primary DeepSeek-blind-judge protocol. A Gemini-2.5-Flash independent-judge rescore of the same 40 baseline packages preserves the system-level ranking, with FirstResearch scoring 4.86/5 versus 4.38/5 for the strongest baseline and Pearson agreement of 0.865 on average score. A one-repeat ablation checkpoint further suggests that the certificate-centered core is the strongest component: certificate-only scoring reaches 4.90/5 under DeepSeek and 4.88/5 under Gemini, while removing certificates drops below 1/5 under both judges. These results are preliminary and use LLM judges rather than human domain experts, but they support a narrow scientific-discovery claim: explicit derivation constraints are a promising mechanism for making LLM-generated scientific questions more auditable. Code, prompts, saved outputs, and reproduction scripts are available at https://github.com/louiswang524/FirstResearch.

cs.AI↗

arXiv2026

When Should Forecasting Agents Reason? Behavioral Stress Tests for Reliability Routing

Forecasting agents increasingly combine language-model reasoning, retrieval, ensembling, and calibration, but it remains unclear when each behavior should be trusted. We study this question on ForecastBench-style binary forecasting tasks, treating the choice to retrieve, reason, defer to a market prior, or use a historical analog as an observable agent behavior rather than a hidden implementation detail. Our central finding is that mechanism choice is source-dependent: structured analogs dominate for some data-generating processes, while market/crowd-style and conservative baselines are better for others. We introduce ReliabilityRoute, a structural intervention that steers forecasting-agent behavior using reliability features such as historical coverage, market-prior availability, source-prior sharpness, evidence strength, evidence disagreement, and horizon. A fixed 2024-fitted rule closely matches a hand taxonomy without hard-coded source-name decisions, while a walk-forward self-adjusting rule refits thresholds from previously resolved vintages and obtains the best mean Brier score among our deterministic systems across 16 later LLM vintages. The gain is modest and historical/search baselines remain highly competitive. The main contribution is therefore a behavioral stress test showing that more reasoning is not always better; forecasting agents should first estimate which evidence source deserves control, routing policies should themselves adapt under auditable constraints, and reproducibility artifacts are available at https://github.com/louiswang524/forcastagent

cs.AI↗

arXiv2026

Grounding Spatial Relations in a Compact World Model: Instruction Leakage and a Goal-Free Dynamics Fix

Compact world models that condition on a language goal promise to ground relations such as ``put the red block left of the blue block'' using a sparse set of explicit \emph{reference anchors}. We ask when such references actually ground a relation, and identify a trap: a goal-conditioned predictor reaches a striking $0.90$ relation-readout accuracy, yet this is \emph{instruction transcription}, not perception. Withholding the goal collapses it to chance ($0.90\!\to\!0.27$, three seeds) and a counterfactual instruction makes the predicted anchors follow the \emph{false} instruction $94.5\%$ of the time (true scene $2.3\%$; $N{=}256$). Tested across three settings and a within-task ablation, our central claim characterizes the confound: \textbf{instruction leakage occurs when the scored quantity is transcribable from the instruction (when the instruction names the answer) and is essentially independent of how predictive the non-instruction inputs are.} Our tabletop and the external BabyAI benchmark leak, whereas a Language-Table forward-dynamics world model whose instruction names \emph{referents} does not, until the instruction is augmented to name the direction; and degrading the action never increases leakage, the opposite of what predictor-competition predicts. The diagnosis prescribes the fix: keep the goal out of the dynamics (it belongs to the planner's cost) and supervise the \emph{read} path, recovering genuine, instruction-independent grounding ($0.88$, identical with and without the goal). The detection protocol and remedy apply to any goal-conditioned world model whose instruction names the scored quantity.

cs.AI↗

arXiv2026

Doing What They Say, Not What They Reason: Locating the Faithfulness Gap in LLM Agents

Do LLM agents act on the reasoning they state? This question of process fidelity is central to LLM-based social simulation, yet hard to measure where no reference for correct behavior exists. We study it in a controlled setting: a Texas Poker simulator with a verifiable reference action for every decision by splitting the faithfulness gap into two steps: reasoning-to-conclusion (does the stated decision follow from the agent's own reasoning?) and conclusion-to-action (does the agent execute what it states?). The two steps behave very differently. Conclusion-to-action is reliable: inconsistency is 0.7% for Claude Haiku 4.5 and 1.4% for DeepSeek-Reasoner once the conclusion is read from an explicit tag, whereas free-text conclusion extraction reports 22-26%. Reasoning-to-conclusion is where fidelity frays, but not through a single dominant failure. In a step-level diagnostic the agent's errors split roughly evenly between bad inputs, borderline cases, and rule misapplication deriving a conclusion that contradicts the agent's own restated rule from inputs it estimated correctly. This composition is model-dependent: rule misapplication accounts for a third of Haiku's interpretable errors but only 8% of DeepSeek's. The one robust signal is directional: when an agent does misapply its own stated rule, it almost always (99.5% for Haiku) errs in the risk-averse direction. The override is partly hedging behavior, not a capability limit: instructing the agent to apply the rule mechanically halves the misapplication rate (13.9% to 6.8% of decisions) and raises adherence by eight points. Process-fidelity evaluation should therefore elicit machine-checkable conclusions and probe for directional biases rather than assume a single upstream failure mode, lest it conflate measurement noise with model behavior.

cs.AI↗

arXiv2026

The Active Ingredient in Muon's Grokking

The Muon optimizer reaches the grokking threshold on modular arithmetic faster than AdamW. Prior work attributes this to "spectral-norm constraints plus orthogonalized momentum" but does not isolate which mechanism matters. To better understand Moun's behavior, we run multi-seed and multi-learning-rate sweeps to decompose and stress-test the effect. First, an ablation shows the speedup comes from orthogonalization (the Newton-Schulz iteration): orthogonalize-only matches full Muon, whereas spectral-only is no faster than AdamW and is unreliable, and this verdict holds across learning rates. Second, a mechanistic analysis finds that orthogonalizing optimizers reach generalization at roughly 3x lower spectral norm and, controlling for how much the embedding actually moves, settle into a lower-norm solution rather than simply perturbing the embedding less. Third, reducing the Newton-Schulz iteration count from five to one accelerates reaching the threshold but makes the grokked solution fragile, prone to transient collapse, with fragility that grows with learning rate; a single iteration is fast and stable only at small learning rate, while the canonical five iterations are the learning-rate-robust choice. We also show spectral scaling can be dropped at no measured cost. A methodological thread runs throughout: under a stability-aware metric, "faster" claims about grokking optimizers can invert, so we report both first-crossing and sustained-grok times. To support reproducibility, we release our full training and analysis code at https://github.com/louiswang524/muon-grokking-frontier

cs.LG↗

arXiv2026

Zero-Parameter Geometric Gating for Temporally Stable Low-Altitude UAV Video Semantic Segmentation

Video semantic segmentation for low-altitude UAVs requires temporal consistency, yet dense optical flow introduces spatially structured noise in the planar regions that dominate aerial imagery. We propose a zero-parameter geometric gate that uses RANSAC homography inlier ratios on a $16\times16$ spatial grid to route each region to either homography or optical flow warp before fusion via Semantic Similarity Propagation. The gate requires no learned parameters -- only a median-threshold binary decision on RANSAC statistics -- adding only 211K trainable parameters (the SSP fusion layer) to a frozen backbone. On synthetic UAVid, the method achieves +4.24--4.91\% mIoU improvement over base models across two architectures (SegFormer-b2 and Hiera-S+UPerNet). Mechanism diagnostics reveal that flow residuals in planar regions are spatially autocorrelated (Moran's I = 0.32, $p < 0.001$), predict boundary instability (Spearman $ρ= 0.66$), and that rigidification recovers temporal consistency from 62\% to 92\% (+29.5pp) in homography-valid regions.

cs.CV↗

arXiv2026

AtlasGS: Brain MRI Spatial Resolution Harmonization With Shared Gaussian Geometry

Splatting (GS)-based shared geometry framework adopts a two-stage training strategy, in which an explicit, subject-specific Gaussian scaffold encoding anatomical geometry is first learned from the isotropic structural scan and then reused to fit appearance for target modalities acquired with sparse slices. Experiments on the UK Biobank, GBM, and ABCD datasets for through-plane super-resolution across multiple modalities (T2-weighted, FLAIR, DWI, ASL), degradation factors ($\times 3$, $\times 5$, $\times 7$), and pathological abnormalities (glioblastoma) demonstrate state-of-the-art reconstruction fidelity. The shared Gaussian geometry enables arbitrary-view generation for target modalities with strong structural consistency and further shows potential for self-supervised in-plane super-resolution. This work establishes explicit geometry-guided representations as a novel, flexible, and interpretable pathway toward retrospective multi-contrast MRI harmonization and reliable clinical reference construction. Source code is available at: https://github.com/yfgao76/AtlasGS

eess.IV↗

arXiv2026

Conservative Discrete Structure Stabilizes Autoregressive Rollouts in a 1D Drift Diffusion Poisson Benchmark

Learned plasma transport surrogates can match short horizon states while failing over long rollouts because charge accounting, density admissibility, and Poisson compatible field reconstruction are not enforced. We study this issue in a controlled nondimensional one dimensional drift diffusion Poisson benchmark with Dirichlet electrostatic potential boundaries and zero species wall fluxes. The benchmark is a conservation and rollout test, not a complete sheath wall model. We compare Conservative FluxNet, a structure preserving flux correction model with a conservative finite volume update and positivity aware limiting, against direct next state regressors, direct variants with Poisson recomputation, charge projection, and rollout training, and a classical conservative core without learned correction. The central result is that the classical finite volume core alone achieves near roundoff rollout error, so the paper is primarily about conservative discrete structure rather than learned closure. On the headline experiment, the conservative model achieves rollout MSE $7.35\times 10^{-9}$ versus $4.23\times 10^{1}$ for the unconstrained baseline, $2.53\times 10^{1}$ with Poisson recomputation, $6.72\times 10^{1}$ with charge projection, and $2.71\times 10^{1}$ with four step rollout training. Across $64$ prespecified configurations, it wins rollout mean squared error in $60/64$ cases despite winning one step mean squared error in only $19/64$. These results show that, for this controlled benchmark and comparison class, local conservative finite volume structure is more important than one step neural regression accuracy for stable autoregressive rollout.

cs.CE↗

arXiv2026

SWE-Mutation: Can LLMs Generate Reliable Test Suites in Software Engineering?

Evaluating software engineering capabilities has become a core component of modern large language models (LLMs); however, the key bottleneck hindering further scaling lies not in the scarcity of high-quality solutions, but in the lack of high-quality test suites. Test suites are indispensable both for synthesizing program repair trajectories and for providing precise feedback signals in reinforcement learning. Unfortunately, due to the high cost and difficulty of annotation, high-quality test suites have long been hard to obtain, while those automatically generated by LLMs tend to be superficial and lack sufficient discriminative power. As a first step toward constructing high-quality test suites, we introduce SWE-Mutation, a benchmark for evaluating LLM-generated test suites. The benchmark characterizes test suites by introducing systematically mutated solutions that attempt to ``fool'' the test suites and pass validation. We further propose an agentic, language-agnostic framework for automatically generating complex mutants. Our benchmark consists of 2,636 mutated variants derived from 800 original instances and includes a multilingual subset spanning nine programming languages. Experiments on seven LLMs reveal that even DeepSeek-V3.1 achieves only 10.20% verification and 36.15% detection rates, highlighting the inadequacy of current LLMs. Additionally, our agentic mutation strategy enhances realism, reducing average detection rates from 71.04% to 39.81% compared to conventional methods. These findings expose persistent deficiencies in the ability of current LLMs to generate reliable and discriminative test suites.

cs.SE↗

arXiv2026

Retrieval as a Decision: Training-Free Adaptive Gating for Efficient RAG

Retrieval-Augmented Generation (RAG) improves factuality but retrieving for every query often hurts quality while inflating tokens and latency. We propose Training-free Adaptive Retrieval Gating (TARG), a single-shot policy that decides when to retrieve using only a short, no-context draft from the base model. From the draft's prefix logits, TARG computes lightweight uncertainty scores-mean token entropy, a margin signal derived from the top-1/top-2 logit gap via a monotone link, or small-N variance across a handful of stochastic prefixes-and triggers retrieval only when the score exceeds a threshold. The gate is model-agnostic, adds only tens to hundreds of draft tokens, and requires no additional training or auxiliary heads. On five QA benchmarks spanning short-answer (NQ-Open, TriviaQA, PopQA), multi-hop (MuSiQue), and long-form (ASQA) tasks, TARG consistently pushes the accuracy-efficiency frontier: compared with Alway-RAG, TARG matches or improves EM/F1 while reducing retrieval by 70-90% and cutting end-to-end latency, and it remains close to Never-RAG in overhead. A central empirical finding is that under modern instruction-tuned LLMs the margin signal is a robust default (entropy compresses as backbones sharpen), with small-N variance offering a conservative, budget-first alternative. We provide ablations over gate type and prefix length and use a $Δ$-latency view to make budget trade-offs explicit.

cs.CL↗

arXiv2026

Dropping Anchor and Spherical Harmonics for Sparse-view Gaussian Splatting

Recent 3D Gaussian Splatting (3DGS) Dropout methods address overfitting under sparse-view conditions by randomly nullifying Gaussian opacities. However, we identify a neighbor compensation effect in these approaches: dropped Gaussians are often compensated by their neighbors, weakening the intended regularization. Moreover, these methods overlook the contribution of high-degree spherical harmonic coefficients (SH) to overfitting. To address these issues, we propose DropAnSH-GS, a novel anchor-based Dropout strategy. Rather than dropping Gaussians independently, our method randomly selects certain Gaussians as anchors and simultaneously removes their spatial neighbors. This effectively disrupts local redundancies near anchors and encourages the model to learn more robust, globally informed representations. Furthermore, we extend the Dropout to color attributes by randomly dropping higher-degree SH to concentrate appearance information in lower-degree SH. This strategy further mitigates overfitting and enables flexible post-training model compression via SH truncation. Experimental results demonstrate that DropAnSH-GS substantially outperforms existing Dropout methods with negligible computational overhead, and can be readily integrated into various 3DGS variants to enhance their performances. Project Website: https://sk-fun.fun/DropAnSH-GS

cs.CV↗

arXiv2026

How Do We Research Human-Robot Interaction in the Age of Large Language Models? A Systematic Review

Advances in large language models (LLMs) are profoundly reshaping the field of human-robot interaction (HRI). While prior work has highlighted the technical potential of LLMs, few studies have systematically examined their human-centered impact (e.g., human-oriented understanding, user modeling, and levels of autonomy), making it difficult to consolidate emerging challenges in LLM-driven HRI systems. Therefore, we conducted a systematic literature search following the PRISMA guideline, identifying 86 articles that met our inclusion criteria. Our findings reveal that: (1) LLMs are transforming the fundamentals of HRI by reshaping how robots sense context, generate socially grounded interactions, and maintain continuous alignment with human needs in embodied settings; and (2) current research is largely exploratory, with different studies focusing on different facets of LLM-driven HRI, resulting in wide-ranging choices of experimental setups, study methods, and evaluation metrics. Finally, we identify key design considerations and challenges, offering a coherent overview and guidelines for future research at the intersection of LLMs and HRI.

cs.RO↗

arXiv2026

Beyond Fast and Slow: Cognitive-Inspired Elastic Reasoning for Large Language Models

Large language models (LLMs) have demonstrated impressive performance across various language tasks. However, existing LLM reasoning strategies mainly rely on the LLM itself with fast or slow mode (like o1 thinking) and thus struggle to balance reasoning efficiency and accuracy across queries of varying difficulties. In this paper, we propose Cognitive-Inspired Elastic Reasoning (CogER), a framework inspired by human hierarchical reasoning that dynamically selects the most suitable reasoning strategy for each query. Specifically, CogER first assesses the complexity of incoming queries and assigns them to one of several predefined levels, each corresponding to a tailored processing strategy, thereby addressing the challenge of unobservable query difficulty. To achieve automatic strategy selection, we model the process as a Markov Decision Process and train a CogER-Agent using reinforcement learning. The agent is guided by a reward function that balances solution quality and computational cost, ensuring resource-efficient reasoning. Moreover, for queries requiring external tools, we introduce Cognitive Tool-Assisted Reasoning, which enables the LLM to autonomously invoke external tools within its chain-of-thought. Extensive experiments demonstrate that CogER outperforms state-of-the-art Test-Time scaling methods, achieving at least a 13% relative improvement in average exact match on In-Domain tasks and an 8% relative gain on Out-of-Domain tasks.

cs.AI↗

arXiv2026

Beyond Model Scaling: Test-Time Intervention for Efficient Deep Reasoning

Large Reasoning Models (LRMs) excel at multi-step reasoning but often suffer from inefficient reasoning processes like overthinking and overshoot, where excessive or misdirected reasoning increases computational cost and degrades performance. Existing efficient reasoning methods operate in a closed-loop manner, lacking mechanisms for external intervention to guide the reasoning process. To address this, we propose Think-with-Me, a novel test-time interactive reasoning paradigm that introduces external feedback intervention into the reasoning process. Our key insights are that transitional conjunctions serve as natural points for intervention, signaling phases of self-validation or exploration and using transitional words appropriately to prolong the reasoning enhances performance, while excessive use affects performance. Building on these insights, Think-with-Me pauses reasoning at these points for external feedback, adaptively extending or terminating reasoning to reduce redundancy while preserving accuracy. The feedback is generated via a multi-criteria evaluation (rationality and completeness) and comes from either human or LLM proxies. We train the target model using Group Relative Policy Optimization (GRPO) to adapt to this interactive mode. Experiments show that Think-with-Me achieves a superior balance between accuracy and reasoning length under limited context windows. On AIME24, Think-with-Me outperforms QwQ-32B by 7.19% in accuracy while reducing average reasoning length by 81% under an 8K window. The paradigm also benefits security and creative tasks.

cs.AI↗

arXiv2026

How well can off-the-shelf LLMs elucidate molecular structures from mass spectra using chain-of-thought reasoning?

Mass spectrometry (MS) is a powerful analytical technique for identifying small molecules, yet determining complete molecular structures directly from tandem mass spectra (MS/MS) remains a long-standing challenge due to complex fragmentation patterns and the vast diversity of chemical space. Recent progress in large language models (LLMs) has shown promise for reasoning-intensive scientific tasks, but their capability for chemical interpretation is still unclear. In this work, we introduce a Chain-of-Thought (CoT) prompting framework and benchmark that evaluate how LLMs reason about mass spectral data to predict molecular structures. We formalize expert chemists' reasoning steps-such as double bond equivalent (DBE) analysis, neutral loss identification, and fragment assembly-into structured prompts and assess multiple state-of-the-art LLMs (Claude-3.5-Sonnet, GPT-4o-mini, and Llama-3 series) in a zero-shot setting using the MassSpecGym dataset. Our evaluation across metrics of SMILES validity, formula consistency, and structural similarity reveals that while LLMs can produce syntactically valid and partially plausible structures, they fail to achieve chemical accuracy or link reasoning to correct molecular predictions. These findings highlight both the interpretive potential and the current limitations of LLM-based reasoning for molecular elucidation, providing a foundation for future work that combines domain knowledge and reinforcement learning to achieve chemically grounded AI reasoning.

cs.CL↗