Searcharxiv⌕ Search

arXiv subjects

Kaiwen Shi

Publications and source records attributed to Kaiwen Shi.

At least 19 recordsLinked to original sources

The Average Patient: Do Simulated Psychiatric Patients Answer Like Real People?

Language models are used as simulated psychiatric patients to train clinicians and evaluate AI therapists, and are judged by plausibility or total scores. We give a framework for simulated respondents: an information budget bounds how much simulated patients can differ, and an average-patient hypothesis, that a model answers each item at the level typical of the named condition and severity, implies little variation, a full profile of the condition's typical symptoms and totals at or above the top of the stated severity band. We examine these implications at matched totals on item-level data from eight human samples, including a clinical and an instructed-feigning sample. Across 17 models, simulated patients vary 1.5 to 11 times less than real respondents with the same total and on most questionnaires endorse nearly every symptom at a middling level; on the PHQ-9 their symptoms covary less like real ones. People, honest or feigning, answer "not at all" about as often as genuine patients with the same totals, and more often than every model. Single answers are highly concentrated, and prompt changes, published personas, larger models and reasoning reduce the pattern without removing it. In interviews, patients told a severity are rated as having nearly every symptom, so an assessment agent's recognition of absent symptoms goes untested, and scored against the stated severity the agent places a quarter of them in their band, against about half of patients given a real symptom profile. Having the model word such a profile restores half to two thirds of the variation.

cs.CL↗

BioBlobs: Unsupervised Discovery of Functional Substructures

Protein function is often localized to cohesive substructures, such as catalytic triads, binding pockets, and structural motifs, that occupy only a small fraction of a protein's residues. Yet existing predictive pipelines typically aggregate residue-level embeddings and do not explicitly model or infer these functional substructures, leaving the central biological question unanswered: which substructure of a protein is responsible for its function? We introduce BioBlobs, an encoder-agnostic, end-to-end differentiable framework that compresses a protein into a small set of cohesive substructures (blobs) and predicts function from these blobs alone, so that each blob corresponds to a candidate functional region. Across diverse protein function prediction tasks and both sequence-only and structure-aware protein language models, BioBlobs is competitive with strong baselines while operating on only a small fraction of residues. Trained only on protein function labels, BioBlobs recovers expert-curated functional sites in InterPro and experimentally annotated catalytic sites in the Mechanism and Catalytic Site Atlas, demonstrating unsupervised functional substructure discovery and opening a path to large-scale functional site discovery in unannotated proteins. The discovered blobs adapt their spatial scale to the task, ranging from local catalytic sites to entire structural domains. Code is available at https://github.com/OliverLaboratory/BioBlobs.

q-bio.BM↗

Beyond Correct Answers: How Compositional Cues Shape Visual Judgments

Visual question answering requires models to assign attributes and relations to the correct objects, a problem known as visual binding. Structured prompts make these assignments explicit through labeled fields. Yet an improvement from the full prompt does not reveal which labels help or how they interact. First, we isolate field-label effects across ten models and six tasks by holding the image and claim fixed. These comparisons reveal differences that overall accuracy can hide. With a relation label present, adding object-role labels helps Qwen3 but hurts Qwen2 on Spatial. Second, we compare auxiliary hints that preserve or swap the question's role assignments against a no-hint baseline. Across two image sets, Qwen3's advantage with preserving hints mainly reflects errors caused by swapped hints. Preserving hints remain near baseline. Field order further affects the harm. Third, we test clearer hint wording and direct selection of the relation shown in the image. Clearer hints improve text matching but do not consistently improve visual judgments. In contrast, relation selection partly reduces errors for some hints in Qwen3 and GPT-4.1-mini.

cs.CV↗

Not Every User Says It All: Training Tool-Using Agents Beyond Ideal Simulated Users

Tool-using agents are trained and evaluated almost exclusively with strong simulated users who state every detail clearly, yet real users are often weaker. We construct weak users by masking sentences of a simulated user's replies, holding the task and the user's knowledge fixed, and cross-evaluate agents trained with strong and weak users on tau2-bench and ToolWOZ. Weak users lower the success of every policy, and training with them does not help: no policy trained with weak users, through fixed masks, mixtures or a curriculum, serves weak users better than the strong-trained policy beyond run-to-run noise, and prompting the agent to ask more does not help either. We propose StaGPO (State-Grouped Policy Optimization): the agent writes what the task needs, what is missing and how it is being filled, and GRPO uses this state to compare turns of different rollouts that reach the same situation, assigning step-level credit. Within the standard recipe of a small SFT cold start followed by RL, StaGPO makes weak-user training useful: on ToolWOZ, the Qwen3.5-2B policy trained with weak users reaches 76.8% with weak users, the highest of any Qwen3.5-2B policy we trained, 4.8 points above GRPO and 2.8 above the strongest baseline; it is also 5.4 points above GRPO with strong users, and generalises to an unseen mask rate; trained with strong users instead, it serves weak users at least as well as GRPO trained with weak users. On tau2-bench it scores highest with weak users for both models, by a small margin for Qwen3.5-4B.

cs.AI↗

GradAgent: A Knowledge-Guided Multi-Agent System for Structure-Preserving Gradient-Flow Computation with an Application to Multicomponent Vesicle Dynamics

High-order differential operators and nonlinear coupling make it challenging to construct conservative and energy-stable schemes for coupled gradient-flow systems. We present GradAgent, a knowledge-guided multi-agent system that coordinates three agents across model analysis, algorithm design and proofs, and numerical implementation and validation. Independent audits strengthen reliability by uncovering mathematical errors and proof gaps, guiding revisions, and maintaining consistency across stages. In Reconstruction Mode, GradAgent reconstructs 20 published studies and organizes audited knowledge in an extensible knowledge graph (KG), GradAgent-KG, linking model structures, discretization strategies and proofs, implementations, and numerical evidence. In Design Mode, the agents assess the applicability of retrieved knowledge and develop new schemes informed by relevant discretization strategies. Applied to the fully coupled multicomponent vesicle phase-field-fluid model, GradAgent yields three first-order and three second-order schemes across three algorithmic families, including four linear, decoupled schemes. Under stated assumptions, all six schemes conserve membrane component mass and vesicle volume and dissipate their respective temporally discrete energies unconditionally. Comparisons with and without GradAgent-KG show that it promotes diversity in structure-preserving scheme design for this target model. Numerical tests confirm second-order spatial accuracy, the expected temporal orders, conservation, and temporally discrete energy dissipation, while three-dimensional shear-flow simulations agree qualitatively with experiments. These results demonstrate GradAgent's ability to combine reusable knowledge, coordinated reasoning, and independent auditing to develop and validate structure-preserving algorithms for complex coupled systems.

math.NA↗

HoosierHelp: Benchmarking LLM Agents for Social Service Navigation

Social service navigation requires connecting help-seeking individuals to resources that satisfy their needs and specific constraints. Although LLM agents offer a promising interface for conversational resource navigation, existing benchmarks do not capture the interaction complexity and constraint-grounding demands of this setting. We introduce HoosierHelp, an interactive benchmark grounded in 3,971 Indiana public social service resources. Agents interact with simulated users, issue structured resource-search calls, handle non-ideal interactions, and select the final resources returned by the tool. HoosierHelp enhances the realism of simulated users by varying their need structure, constraint satisfiability, and behavior patterns, including impatience, rambling, unsupported requests, and self-contradiction. Experiments on 240 samples across seven LLMs show that current LLM agents remain substantially unreliable for social service navigation. Performance drops sharply on fallback-required and self-contradictory conversations, highlighting the need for agents that are more robust to complex and non-ideal user interactions.

cs.HC↗

Policy4OOD: A Knowledge-Guided World Model for Policy Intervention Simulation against the Opioid Overdose Crisis

The opioid epidemic remains one of the most severe public health crises in the United States, yet evaluating policy interventions before implementation is difficult: multiple policies interact within a dynamic system where targeting one risk pathway may inadvertently amplify another. We argue that effective opioid policy evaluation requires three capabilities -- forecasting future outcomes under current policies, counterfactual reasoning about alternative past decisions, and optimization over candidate interventions -- and propose to unify them through world modeling. We introduce Policy4OOD, a knowledge-guided spatio-temporal world model that addresses three core challenges: what policies prescribe, where effects manifest, and when effects unfold.Policy4OOD jointly encodes policy knowledge graphs, state-level spatial dependencies, and socioeconomic time series into a policy-conditioned Transformer that forecasts future opioid outcomes.Once trained, the world model serves as a simulator: forecasting requires only a forward pass, counterfactual analysis substitutes alternative policy encodings in the historical sequence, and policy optimization employs Monte Carlo Tree Search over the learned simulator. To support this framework, we construct a state-level monthly dataset (2019--2024) integrating opioid mortality, socioeconomic indicators, and structured policy encodings. Experiments demonstrate that spatial dependencies and structured policy knowledge significantly improve forecasting accuracy, validating each architectural component and the potential of world modeling for data-driven public health decision support. Our code and data have been released in https://github.com/antman9914/Policy4OOD.

cs.LG↗

Food4All: An Agentic Framework and Benchmark for Food Resource Navigation with Adaptive User Understanding

Food assistance referral requires conversational agents to translate underspecified, often noisy help-seeking dialogues into locally valid resource recommendations. We present Food4All, an agentic food-resource referral framework and benchmark grounded in 686 structured Indiana food resources. Food4All couples a food-specific search tool with 300 multi-turn evaluation tasks spanning single food needs, composite cases with access or document constraints, and five non-ideal user interaction traits: unreasonable demands, rambling responses, impatience, incomplete answers, and inconsistent information. We evaluate six Large Language Models (LLMs) on requirement grounding, resource retrieval, final referral correctness, and interaction efficiency. Although the strongest model achieves 96.33% referral accuracy, our diagnostics reveal persistent failures in grounding schedule, eligibility, intake, and document constraints, as well as failures to preserve valid retrieved resources in the final recommendation. Trait-level analysis further shows that different non-ideal behaviors stress different parts of the referral pipeline. Food4All provides a controlled testbed for studying tool-calling agents in constraint-sensitive food assistance referral under realistic user interaction challenges.

cs.CL↗

SAGE: Answer-Conditioned Uncertainty Targets for Verbal Uncertainty Alignment

Large language models increasingly express uncertainty through natural-language statements, yet these expressions often fail to reflect the model's sampled behavior. We study verbal uncertainty alignment as a distributional calibration problem: the appropriate uncertainty target for a prompt should be estimated from repeated model outputs rather than from an isolated response. However, group rollouts alone are insufficient, since the resulting target must provide a useful training signal. Existing targets only partially satisfy this requirement. We propose SAGE, Semantic-Answer Guided Entropy, a group-level uncertainty target that constructs an answer-conditioned uncertainty geometry over sampled responses. SAGE preserves categorical, numeric, and symbolic answer distinctions while maintaining a smooth and scale-preserving calibration signal. We further apply this target through Group-Uncertainty Preference Optimization, or GUPO, an uncertainty-channel training framework that supervises verbal uncertainty expressions rather than the full response. Experiments across factual, mathematical, and multiple-choice reasoning tasks show improved uncertainty ranking, lower calibration error, and reduced overconfidence.

cs.CL↗

Confidence Laundering in Agent Systems: Why Uncertainty Needs a Latent Carrier

Modern agent systems can turn uncertainty into overconfidence. Fragile upstream decisions are often exposed to downstream components as clean intermediate artifacts, while the uncertainty behind those decisions is lost at the interface. As a result, local ambiguity can become system-level error amplification. We argue that this reveals an interface bottleneck in agent uncertainty propagation: uncertainty does not propagate simply because a trajectory contains uncertain steps; it propagates only when it survives the handoff between components. We define uncertain decision handoff as the transfer of an intermediate decision made under uncertainty, and identify confidence laundering as a failure mode in which fragile upstream states are repackaged as procedurally valid artifacts that downstream agents over-trust. To address this bottleneck, we propose latent uncertainty as an uncertainty-bearing carrier attached to decision handoffs. Rather than replacing text with hidden states, latent uncertainty aims to preserve pre-commitment fragility in a form that downstream components can use. This position shifts agent uncertainty propagation from step-wise uncertainty estimation toward uncertainty-preserving interface design for more recoverable agent systems.

cs.AI↗

Why Semantic Entropy Fails: Geometry-Aware and Calibrated Uncertainty for Policy Optimization

Post-training has become central to improving reasoning and alignment in large language models, where critic-free models enable scalable learning from model-generated outputs but lack principled mechanisms to distinguish informative from noisy signals. Recent approaches leverage response-level measures as uncertainty signals to regulate group-based optimization methods such as GRPO. Yet their empirical success remains unstable and unclear in how they influence optimization dynamics. In this paper, we provide, to our knowledge, the first principled formulation that interprets uncertainty signals as mechanisms for characterizing and regulating gradient variance and learning signal quality. Based on both empirical and theoretical analysis, we identify two critical gaps of current entropy-based estimators: The anisotropic gap and The calibration gap. Motivated by this analysis, we propose Geometric-aware Calibrated Policy Optimization (GCPO), a novel framework integrating geometry-aware measures to capture semantic disagreement with reward-based calibration to align uncertainty with learning signal strength. Experiments on multiple benchmarks show that our approach more faithfully tracks gradient variability and consistently improves post-training performance. Our results highlight the importance of designing uncertainty signals that are aligned with optimization dynamics, offering a principled perspective for robust post-training.

cs.LG↗

ENSEMBITS: an alphabet of protein conformational ensembles

Protein structure tokenizers (PSTs) are workhorses in protein language modeling, function prediction, and evolutionary analysis. However, existing PSTs only capture local geometry of static structures, and miss the correlated motions and alternative conformational states revealed by protein ensembles. Here we introduce Ensembits, the first tokenizer of protein conformational ensembles. Ensembits address challenges inherent to tokenizing dynamics: deriving informative geometric descriptors across conformations, permutation-invariance encoding of variable-size ensembles, and conquering sparsity in dynamics data. Trained with a Residual VQ-VAE using a frame distillation objective on a large molecular dynamics corpus, Ensembits outperforms all related methods on RMSF prediction, and is the strongest standalone structural tokenizer on an token-conditioned ANOVA test on per-residue motion amplitude. Ensembits further matches or exceeds static tokenizers on EC, GO, binding site/affinity prediction, and zero-shot mutation-effect prediction despite using far less pretraining data. Notably, the distillation objective enables Ensembits to predict dynamics token from one single predicted structure, which alleviates dynamics data sparsity. As the field moves from static structure prediction toward ensemble generation, Ensembits offer the discrete vocabulary needed to bring dynamics into protein language modeling and design.

cs.LG↗

CiteAudit: You Cited It, But Did You Read It? A Benchmark for Verifying Scientific References in the LLM Era

Scientific research relies on citation integrity, yet large language models (LLMs) have introduced a critical risk: fabricated references that appear plausible but correspond to no real publications. As manual verification becomes infeasible and existing automated tools remain fragile, we introduce CiteAudit, a comprehensive benchmark and detection framework for hallucinated citations. We design a multi-agent verification pipeline that decomposes citation checking into metadata extraction, memory lookup, web-based retrieval, and final judgment. To evaluate this, we construct a large-scale, human-validated dataset spanning diverse domains and hallucination types. Experiments demonstrate that our framework achieves superior verification performance over state-of-the-art LLMs and commercial baselines. Our work provides the necessary infrastructure to audit citations at scale and safeguard the trustworthiness of scholarly discourse. Code is available at https://github.com/shiiiikw/CiteAudit.

cs.CL↗

EvolveRouter: Co-Evolving Routing and Prompt for Multi-Agent Question Answering

Large language model agents often exhibit complementary strengths, making routing a promising approach for multi-agent question answering. However, existing routing methods remain limited in two important ways: they typically optimize over a fixed pool of agents without improving the agents themselves, and they often rely on rigid collaboration schemes that cannot adapt the number of participating agents to the query. We propose EvolveRouter, a trainable framework that addresses both limitations by jointly improving agent quality and collaboration structure. First, EvolveRouter couples graph-based query routing with targeted instruction refinement in a closed-loop co-evolution process, allowing router diagnostics to guide agent improvement while refined agents provide cleaner supervision for routing. Second, it introduces an adaptive inference strategy that dynamically determines the effective collaboration size for each query through router-weighted answer agreement. Together, these designs enable more capable and more efficient multi-agent reasoning. Experiments on five question answering benchmarks show that EvolveRouter consistently outperforms SOTA routing baselines in both F1 and exact match, while further analysis confirms the benefits of closed-loop refinement and adaptive collaboration.

cs.CL↗

Scaling Teams or Scaling Time? Memory Enabled Lifelong Learning in LLM Multi-Agent Systems

Large language model (LLM) multi-agent systems can scale along two distinct dimensions: by increasing the number of agents and by improving through accumulated experience over time. Although prior work has studied these dimensions separately, their interaction under realistic cost constraints remains unclear. In this paper, we introduce a conceptual scaling view of multi-agent systems that jointly considers team size and lifelong learning ability, and we study how memory design shares this landscape. To this end, we propose \textbf{LLMA-Mem}, a lifelong memory framework for LLM multi-agent systems under flexible memory topologies. We evaluate LLMA-Mem on \textsc{MultiAgentBench} across coding, research, and database environments. Empirically, LLMA-Mem consistently improves long-horizon performance over baselines while reducing cost. Our analysis further reveals a non-monotonic scaling landscape: larger teams do not always produce better long-term performance, and smaller teams can outperform larger ones when memory better supports the reuse of experience. These findings position memory design as a practical path for scaling multi-agent systems more effectively and more efficiently over time.

cs.MA↗

Drift-Bench: Diagnosing Cooperative Breakdowns in LLM Agents under Input Faults via Multi-Turn Interaction

As Large Language Models transition to autonomous agents, user inputs frequently violate cooperative assumptions (e.g., implicit intent, missing parameters, false presuppositions, or ambiguous expressions), creating execution risks that text-only evaluations do not capture. Existing benchmarks typically assume well-specified instructions or restrict evaluation to text-only, single-turn clarification, and thus do not measure multi-turn disambiguation under grounded execution risk. We introduce \textbf{Drift-Bench}, the first diagnostic benchmark that evaluates agentic pragmatics under input faults through multi-turn clarification across state-oriented and service-oriented execution environments. Grounded in classical theories of communication, \textbf{Drift-Bench} provides a unified taxonomy of cooperative breakdowns and employs a persona-driven user simulator with the \textbf{Rise} evaluation protocol. Experiments show substantial performance drops under these faults, with clarification effectiveness varying across user personas and fault types. \MethodName bridges clarification research and agent safety evaluation, enabling systematic diagnosis of failures that can lead to unsafe executions.

cs.AI↗

FormulaReasoning: A Dataset for Formula-Based Numerical Reasoning

The application of physics formulas is a fundamental human capability in numerical reasoning. While existing datasets often rely on implicit mathematical knowledge, they rarely explicitate the underlying formulas. To address this, we introduce FormulaReasoning, a new benchmark for formula-based numerical reasoning comprising 5,324 questions requiring calculations grounded in external physics principles. We provide high-quality, fine-grained annotations in English and Chinese--including formula structures, parameter names, symbols, values, and units--curated through manual effort and LLM-assisted validation. Additionally, we provide a consolidated formula database as an external knowledge source. To further challenge model performance, we develop an extended version of the dataset by coupling multiple questions. We evaluate various architectural and methodological frameworks, including retrieval-augmented methods, modular reasoning (formula generation, parameter extraction, and calculation), and preference-based optimization. Our analysis identifies critical challenges in formula-based reasoning, highlighting significant opportunities for future methodological advancement.

cs.CL↗

BLURR: A Boosted Low-Resource Inference for Vision-Language-Action Models

Vision-language-action (VLA) models enable impressive zero shot manipulation, but their inference stacks are often too heavy for responsive web demos or high frequency robot control on commodity GPUs. We present BLURR, a lightweight inference wrapper that can be plugged into existing VLA controllers without retraining or changing model checkpoints. Instantiated on the pi-zero VLA controller, BLURR keeps the original observation interfaces and accelerates control by combining an instruction prefix key value cache, mixed precision execution, and a single step rollout schedule that reduces per step computation. In our SimplerEnv based evaluation, BLURR maintains task success rates comparable to the original controller while significantly lowering effective FLOPs and wall clock latency. We also build an interactive web demo that allows users to switch between controllers and toggle inference options in real time while watching manipulation episodes. This highlights BLURR as a practical approach for deploying modern VLA policies under tight compute budgets.

cs.RO↗