arXiv subjects

Zewen Liu

Publications and source records attributed to Zewen Liu.

At least 19 recordsLinked to original sources

arXiv2026

Evo-Harness: Context-to-Harness Skill Compilation for Self-Evolving Agents

Learning from experience is critical for developing capable, self-improving large language model (LLM) agents. Existing methods typically extract knowledge from accumulated trajectories via reflection, memory, rules, or skills. However, agents in realistic environments continuously encounter novel tasks, often offering only a one-shot opportunity to improve. These executions yield rich but highly noisy contexts, entangling broadly useful lessons with task-specific artifacts. Critically, prior works rarely validate their effectiveness on complex real-world tasks or isolate the underlying drivers of improvement. To address these gaps, we formulate online harness learning, where a frozen agent improves by continually updating a structured harness across sequential tasks. This formulation enables a systematic study of key self-improvement factors through our proposed Evo-Harness. At its core, context-to-harness skill compilation distills noisy, single-shot executions into reusable skill harnesses for cross-domain and topic-level adaptation. To demonstrate the efficacy of one-shot skill compilation, we evaluate across five realistic benchmarks (TerminalBench2, SWE-bench, CL-Bench, -bench, WebArena-Infinity). Our extensive analysis demonstrates the effectiveness of Evo-Harness and provides a principled understanding of how LLM agents can effectively learn on the fly. Our code is available at https://github.com/A-EVO-Lab/a-evolve/tree/release/evo-harness.

cs.AI↗

arXiv2026

BOUNDARY_SYNC: Measuring Communication-Induced Representational Coupling in Multi-Agent LLM Systems

As large language models (LLMs) are deployed as communicating agents, does inter-agent communication cause outputs to converge? We introduce BOUNDARY_SYNC, a protocol measuring representational coupling via the Coupling Amplification Factor (CAF = JSD_cond / JSD_baseline), where CAF < 1 indicates homogenization and CAF > 1 indicates diversification. In controlled GPT-4o experiments (N=30, ~9,900 API calls), we measure coupling in text and image communication. Key findings: (1) text communication causes significant homogenization (CAF=0.803 [0.740, 0.873], d=1.30, p<0.001), confirmed by no-communication ablation and prompt-perturbation controls; (2) image communication also homogenizes under within-modality baselines (CAF=0.834 [0.811, 0.858]), with comparable proportional effect; (3) group size moderates coupling direction -- K=5 produces homogenization while K=3 yields CAF > 1.0 (point estimates 1.14 and 1.06, CI pending), suggesting a directional shift toward diversification; (4) cross-model replication shows extreme variation (CAF 0.034-0.803), with DeepSeek dominated by format artifacts; (5) coupling is stateless -- driven by prompt context rather than cumulative updating, with continuous consensus producing monotonic convergence. These results establish LLM agent coupling as real, measurable, and controllable at the prompt level, with direct implications for multi-agent system design.

cs.LG↗

arXiv2026

EPC: A Standardized Protocol for Measuring Evaluator Preference Dynamics in LLM Agent Systems

When LLM agents use evaluator feedback to adapt their behavior in closed loops, evaluator biases propagate through the agent's strategy distribution -- a phenomenon known as evaluator preference coupling. Prior work has documented coupling across multiple evaluator families and model versions, but the field lacks a standardized protocol that enables third-party researchers to (i) reproduce coupling measurements, (ii) compare results across evaluators and time points, and (iii) detect measurement decay as proprietary evaluators silently update. This paper provides the protocol. We specify EPC (Evaluator Preference Coupling) -- a detailed, RFC-style protocol specification for the four-phase isolation paradigm, covering executor and evaluator configuration, strategy and task design, the TTRL update rule, metric computation (gamma, JSD, ECE, Brier), and output schema. We accompany the protocol with a versioned Reference Snapshot v1.0: coupling measurements for eight evaluator conditions (N=122 unique experimental repetitions across GPT-4o, Qwen, DeepSeek, and others) derived from five independent studies, annotated with evaluator version identifiers, API endpoints, and measurement dates. The snapshot is explicitly time-bound: all values are conditional on specific model versions and are expected to decay as proprietary evaluators update. We define a versioning convention (vX.Y-Z, encoding protocol version, snapshot version, and evaluator generation) and provide a usage guide covering adoption, interpretation, and known pitfalls. The protocol, reference snapshot, and implementation code are released as open infrastructure.

cs.LG↗

arXiv2026

Mapping the Evaluation Frontier: An Empirical Survey of the Bias-Reliability Tradeoff Across Eleven Evaluator-Agent Conditions

The bias-reliability tradeoff conjectures that LLM evaluation systems are constrained in (gamma, H, CV) space, where evaluator coupling (gamma), strategy diversity (H), and small-sample measurement reliability (CV(N)) cannot be simultaneously optimized at fixed sample size N. Prior evidence rests on n=5 conditions with complete metrics from a single study. We expand the empirical base to 11 conditions, measuring gamma and H for all 11 (nine with valid weight vectors) and CV(N=5) for seven with sufficient seeds (N >= 5). Five conditions provide the complete (gamma, H, CV) triple. The data confirm the trade-off: conditions with low evaluator coupling (gamma < 0.2) exhibit high measurement noise (CV(N=5) > 1.0), while conditions with strong coupling (gamma > 0.9) achieve low noise (CV(N=5) < 0.16). The correlation r(H, gamma) = -0.989 (n=5, excluding GPT-4o conditions) confirms that evaluator coupling suppresses strategy diversity. Four GPT-4o conditions show gamma=0.000 and H=1.000 across all seeds -- a pattern we attribute to version drift in the June 2026 GPT-4o API. No condition occupies the region {gamma < 0.2, CV(N=5) < 0.3}. We release all per-condition metrics as a standardized benchmark dataset for evaluator comparison.

cs.LG↗

arXiv2026

Calibrating the Evaluator: Does Probability Calibration Mitigate Preference Coupling in LLM Agent Feedback Loops?

When large language model (LLM) agents adapt their behavior through evaluator feedback, systematic evaluator biases propagate into the agent's learned strategy distribution - a phenomenon termed evaluator preference coupling. Prior work has documented this coupling and established a diagnostic framework (EPC) to measure it, but has not investigated whether calibration techniques can mitigate the effect. We present the first study of evaluator calibration as mitigation: applying probability calibration to the evaluator's pairwise judgments to reduce spurious preference propagation. In a controlled within-subjects experiment (N=5) comparing standard binary TTRL (win/loss) with confidence-calibrated TTRL (probability-weighted updates) using DeepSeek-V4-Pro as executor and GLM5.2 as evaluator, we find that calibration reduces the coupling coefficient gamma by 20-49% and Jensen-Shannon divergence by 45-67%. A symmetric-LR control confirms the effect is not due to reduced update asymmetry. We release the calibrated TTRL protocol and recommend it as a lightweight mitigation for LLM-as-judge deployment pipelines.

cs.LG↗

arXiv2026

The Hidden Cost of Resampling: How Imbalance Correction Degrades Probability Calibration in Tree Ensembles

Resampling methods such as SMOTE and random under/over-sampling are standard tools for class-imbalanced classification, almost always evaluated by minority-class accuracy or F1. Prior work has established that undersampling degrades probability calibration by distorting the training prior [1]. We extend this lens to synthetic oversampling (SMOTE) and provide a practical, evidence-based guide to when calibration damage matters and how to fix it. Across five public datasets (imbalance ratio 1.9-70) and two ensemble models (random forest, gradient boosting), with ten seeds and paired statistics, we find: (1) SMOTE's calibration cost is real but small (ECE +0.009; Cliff's delta = +0.27, small-to-moderate) across the studied imbalance range (IR 1.9-70) and its discrimination gains typically outweigh the calibration penalty; (2) random undersampling is the genuine danger -- its damage grows sharply with imbalance, inflating ECE from 0.008 to 0.395 on a dataset with ratio 70, largely because the resulting training sets are too small to estimate probabilities reliably; (3) a single post-hoc recalibration step (Platt or isotonic) eliminates the damage, reducing ECE by up to 66% at a negligible ranking-power cost (AUC -0.002, Cliff's delta = -0.07); and (4) the analytic prior-shift correction that repairs undersampling does not transfer to SMOTE, because SMOTE distorts the class-conditional density rather than only the prior -- so data-driven recalibration remains necessary. We recommend that imbalanced-learning studies report calibration alongside discrimination, and that practitioners recalibrate after resampling whenever predicted probabilities drive decisions.

cs.LG↗

arXiv2026

The Contagion Tensor: A Framework for Measuring Output-Distribution Coupling in Multi-Agent LLM Systems -- and Auditing the Claims It Enables

We introduce the Contagion Tensor, a measurement framework for quantifying how large language model (LLM) output distributions couple across modalities, agents, and time steps. From the tensor we derive the Coupling Amplification Factor (CAF), a family of ratio-based metrics sharing the form CAF = E[T_condition] / E[T_baseline], providing unitless, baseline-referenced measurement with bootstrap confidence intervals. We instantiate CAF in four variants and evaluate the strongest in a complete 2x2x2 block-orthogonal simulation design with modality-specific ablation. The ablation reveals that an apparent image-condition super-linear effect (CAF = 1.40) collapses to sub-linear (CAF = 0.87) when the image perturbation module is disabled, a shift of -0.53 with zero effect on text conditions. We supplement with real-API experiments across two model families: DeepSeek-Chat (R=30) and GPT-4o-mini (R=15, real vision). Under uniform personas, text-only communication produces CAF approx 1.0 in both models. Diverse personas drive convergence (CAF = 0.88). A within-model comparison on GPT-4o-mini reveals: C3 (text) CAF = 1.02 vs. C5 (real vision, R=30) CAF = 1.72 [1.700, 1.733], delta = +0.70, validating the simulation's super-linear image-condition prediction. Of 11 conditions, 5 have been tested on real APIs and 6 remain unverified. Our contribution is two-layered: (1) a measurement instrument that makes output-distribution coupling quantitatively falsifiable; and (2) a transferable ablation protocol that any modular multi-agent simulator can adopt to distinguish genuine coupling from design artifacts.

cs.LG↗

arXiv2026

TMP: Tree-structured Mixed-policy Pruning for Large-scale Image Generation and Editing

Modern image generation model rapidly grows their sizes to meet high-fidelity image synthesis. However, they gradually become unaffordable for their enormous parameter consumption and computation budget that lead to massive resources requirement and gpu memory footprint. In this paper, we propose TMP, the first Tree-structured Mixed-policy Pruning framework that generalizes prevalent image tasks (T2I and TI2I) and architectures (Mixture-of-Experts (MoE) and Diffusion transformer (DiT)). It could be applied to the step-distilled models and contribute as the last stage. We perform experiments upon current open-sourced SOTA HunyuanImage-3.0 instruct and a popular efficient model Z-Image turbo. The proposed pruning framework manages to compress HunyuanImage 3.0 from 80B to 20B parameters at 75% reduction ratio, sacrificing limited generation quality. We also optimize to enable the inference of the pruned 20B version of HunyuanImage 3.0 on a single 24GB 4090 GPU by engineering skills. The inference script and model weight have been integrated into the existing HunyuanImage3.0 open-source github and huggingface repository. Besides, we prove the efficacy of TMP by compressing Z-Image turbo from 6B to 4B (33% reduction) with negligible degradation.

cs.CV↗

arXiv2026

Memory Contagion: Cross-Temporal Propagation of Evaluator Bias via Agent Memory

Large Language Model (LLM) agents increasingly rely on memory systems to maintain long-term coherence. Recent work shows that agent memories degrade during continuous consolidation. However, existing research assumes memories are derived from unbiased experiences. In this work, we identify and formalize a novel phenomenon: Memory Contagion -- the cross-temporal propagation of evaluator bias through agent memory. We show that when agents are trained or guided by biased evaluators, their experiences become biased; when these trajectories are stored and consolidated into memory, the bias propagates to future agents retrieving from the same memory store, even when consolidation is perfect (oracle). Across two bias types (length preference, authority bias) and four experimental phases, we demonstrate: (1) Memory Contagion occurs for length bias even with perfect consolidation on older models (Gamma_A = 13.18, DeepSeek V4-Chat), while newer models (V4-Pro, Claude) are immune, proving both that biased input is a sufficient cause and that contagion is model-generation-dependent; (2) authority bias fails to propagate in all 15 controlled multi-seed experiments (Gamma_A = 0.00), revealing that not all evaluator biases can cross temporal boundaries through current memory architectures; (3) No observed safe threshold: length bias propagation is detected at contamination rates as low as p=0.2. Our findings expose a critical but contingent vulnerability in current agent memory designs and provide formal tools for measuring cross-temporal bias propagation.

cs.LG↗

arXiv2026

Contagion Networks: Evaluator Preference Propagation in Multi-Agent LLM Systems

When large language models serve as evaluators in multi-agent systems, their strategy preferences -- whether induced by explicit prompts or by shared architectural priors -- propagate through the agent network. We introduce Contagion Networks, a formal framework for measuring how evaluator preferences spread across interacting LLM agents. In a controlled 3-agent experiment using DeepSeek-chat with three distinct evaluator preference profiles (structured, balanced, evidence-based), we measure the Cross-Agent Contagion Matrix Gamma_3 and find that preferences consistently propagate between agents (gamma in [0.157, 0.352]). A neutral-prompt control experiment reveals a counter-intuitive result: shared architectural priors dominate explicit preference prompts as the driver of contagion (rho_neutral = 1.498 vs. rho_mixed = 1.299; prompt contribution: -63.5%). We identify three propagation regimes governed by the spectral radius rho(Gamma_N) and demonstrate that the same agents suppress preference contagion in chain topology (beta_3 = 0.0126 +/- 0.0038, 95% CI [0.0089, 0.0163], n=4 seeds) but cascade in fully-connected topology (Delta H_avg = -0.020) -- a topology-dependent regime transition validated both for homogeneous and cross-model agent pools (rho^cross = 1.296 +/- 0.016, n=4). We show that increasing evaluator committee size from k=1 to k=3 reduces effective contagion by 68.9% +/- 14.1% (n=4 seeds), providing an actionable mitigation strategy. We release the open-source Contagion Network experimental framework.

cs.LG↗

arXiv2026

Multimodal Evaluator Preference Collapse: Cross-Modal Coupling in Self-Evolving Agents

When AI agents use language models to evaluate their own outputs in a feedback loop, systematic biases emerge. We show that Evaluator Preference Collapse (EPC) is dramatically amplified in multimodal settings. Using GPT-4o to evaluate DeepSeek-chat across text and visual tasks, we find that a single strategy (step_by_step) absorbs 48.4% of all weight -- 3.2x the collapse observed in text-only self-evaluation -- while three visual-domain strategies receive only 9.1% combined weight. We then demonstrate a novel phenomenon we term cross-modal coupling: evaluator preferences acquired on one modality transfer to and corrupt strategy selection on another. Through a four-phase isolation training paradigm, we measure coupling coefficients and document strategy inversion -- the optimal strategy for a modality reverses after cross-modal exposure. A Phase 3 statistical validation across five evaluator configurations (N=80 total independent repetitions, ~35,000 API calls) with both text-proxy and real-image visual tasks finds: cross-model evaluation produces strong coupling (JSD~0.19-0.34), real-image inputs yield the most directionally consistent signal (mean gamma_{T->V}=1.145, gamma_{V->T}=0.937, 70% T->V, Cohen's d=0.56), and self-evaluation provides near-complete immunity -- 97% of runs (N=30) yield zero coupling (JSD=0.003, d=0.07). Three methodological ablations and multi-executor validation confirm the effect is not a structural artifact. We introduce the coupling matrix indexed by evaluator identity, release the MM-EPC framework, and identify cross-model evaluator architecture as the primary risk factor for preference drift. Code and data: https://github.com/aidless/mm-epc.

cs.LG↗

arXiv2026

Adaptive Auto-Harness: Sustained Self-Improvement for Agentic System Deployment on Open-Ended Task Streams

Auto-harness systems such as A-Evolve, GEPA, and Meta-Harness improve LLM agents by optimizing prompts, skills, tools, memories, and supporting infrastructure from execution feedback, but they are typically evaluated on fixed offline benchmarks. Real deployments instead present open-ended task streams: histories grow without a fixed endpoint, heterogeneous tasks require different harnesses, and problem distributions shift over time. These challenges make a single repeatedly and densely updated harness brittle, causing performance degradation as accuracy peaks early and then declines. This motivates sustained harness construction with task-wise adaptation. We introduce Adaptive Auto-Harness, a framework and system for such streams. The framework decomposes the gap to an oracle harness into evolution loss and adaptation loss. The system addresses these losses with a stateful multi-agent evolver, a harness tree with solve-time routing, and human-steering hooks for cases where history lacks the needed signal. Across prediction-market, security-competition, and event-forecasting streams, Adaptive Auto-Harness outperforms five existing auto-harness baselines and ablations attribute gains to better construction, routing, or targeted human steering. Code is available in \href{https://github.com/A-EVO-Lab/a-evolve/tree/release/adaptive-auto-harness}{Link}.

cs.LG↗

arXiv2026

Harness Updating Is Not Harness Benefit: Disentangling Evolution Capabilities in Self-Evolving LLM Agents

LLM agents are increasingly deployed as systems built around editable external harnesses, including prompts, skills, memories and tools, that shape task execution without changing model parameters. Harness self-evolution adapts such agents by updating these harnesses from execution evidence. Yet it remains unclear whether a model's base capability in task-solving predicts its capabilities in harness self-evolution: which models produce useful harness updates, and which actually benefit from them? We analyze two harness self-evolution capabilities: (i) harness-updating, the capability to produce useful persistent harness updates from execution evidence; (ii) harness-benefit, the capability to benefit from updated harnesses during task solving. Our analysis reveals two findings. First, harness-updating is flat in base capability: models from different capability tiers produce harness updates that lead to surprisingly similar gains; even Qwen3.5-9B's updates yield gains comparable to those of Claude Opus~4.6. Second, harness-benefit is non-monotonic in base capability: weak-tier models benefit little from updated harnesses, mid-tier models benefit most, and strong-tier models benefit less than mid-tier. We trace low gains at the weak tier to two failure modes: weak-tier models may fail to activate relevant harness artifacts, or activate them but fail to follow them faithfully. These findings suggest investing capability budget in the task-solving agent rather than the evolver, and targeting harness invocation and long-horizon instruction following in agent training. Our source code is publicly available at https://github.com/A-EVO-Lab/a-evolve/tree/release/harness-evolution.

cs.AI↗

arXiv2026

Development and Performance Study of Vertical GaN $\alpha$-Particle Detector with High Energy Resolution

High-energy-resolution GaN $\alpha$-particle detectors have significant potential for space radiation, nuclear instrumentation, and harsh-environment applications. However, existing GaN $\alpha$-particle detectors still face several key challenges, including reducing the dead-layer thickness, suppressing leakage current under high reverse bias, improving energy resolution, and clarifying the physical mechanism underlying the low-energy tail phenomenon. This study presents a vertical homoepitaxial GaN $\alpha$-particle detector integrating a 20-nm ultrathin dead layer and a guard-ring structure. The detector exhibits an ultralow leakage current of 2.195 nA at -200 V and an intrinsic energy resolution of 2.69% with a charge collection efficiency (CCE) of 95.9% at -260 V. More importantly, this work demonstrates for the first time through Geant4 simulations that depletion-width nonuniformity is the dominant source of partial energy leakage, leading to an extended low-energy tail in the energy spectrum. We establish a depletion-width nonuniformity model and observe good agreement between simulation and experiment. This finding provides practical guidance for the design and optimization of high-performance GaN-based radiation detectors.

physics.ins-det↗

arXiv2026

Exposing Vulnerabilities in Explanation for Time Series Classifiers via Dual-Target Attacks

Interpretable time series deep learning systems are often assessed by checking temporal consistency on explanations, implicitly treating this as evidence of robustness. We show that this assumption can fail: Predictions and explanations can be adversarially decoupled, enabling targeted misclassification while the explanation remains plausible and consistent with a chosen reference rationale. We propose TSEF (Time Series Explanation Fooler), a dual-target attack that jointly manipulates the classifier and explainer outputs. In contrast to single-objective misclassification attacks that disrupt explanation and spread attribution mass broadly, TSEF achieves targeted prediction changes while keeping explanations consistent with the reference. Across multiple datasets and explainer backbones, our results consistently reveal that explanation stability is a misleading proxy for decision robustness and motivate coupling-aware robustness evaluations for trustworthy time series tasks.

cs.LG↗

arXiv2026

EpiQAL: Benchmarking Large Language Models in Epidemiological Question Answering and Reasoning

Reliable epidemiological reasoning requires synthesizing study evidence to infer disease burden, transmission dynamics, and intervention effects at the population level. Existing medical question answering benchmarks primarily emphasize clinical knowledge or patient-level reasoning, yet few systematically evaluate evidence-grounded epidemiological inference. We present EpiQAL, to our knowledge the first diagnostic benchmark for epidemiological question answering over research literature, comprising three subsets built from open-access articles across diverse diseases. The three subsets progressively test factual recall, multi-step inference, and conclusion reconstruction under incomplete information, and are constructed through a quality-controlled pipeline combining taxonomy guidance, multi-model verification, and difficulty screening. Experiments on fifteen models spanning open-source and proprietary systems reveal that current LLMs show limited performance on epidemiological reasoning, with multi-step inference posing the greatest challenge. Model rankings shift across subsets, and scale alone does not predict success. Chain-of-Thought prompting benefits multi-step inference but yields mixed results elsewhere. EpiQAL provides fine-grained diagnostic signals for evidence-grounding, inferential reasoning, and conclusion reconstruction.

cs.CL↗

arXiv2025

Generalist vs Specialist Time Series Foundation Models: Investigating Potential Emergent Behaviors in Assessing Human Health Using PPG Signals

Foundation models are large-scale machine learning models that are pre-trained on massive amounts of data and can be adapted for various downstream tasks. They have been extensively applied to tasks in Natural Language Processing and Computer Vision with models such as GPT, BERT, and CLIP. They are now also increasingly gaining attention in time-series analysis, particularly for physiological sensing. However, most time series foundation models are specialist models - with data in pre-training and testing of the same type, such as Electrocardiogram, Electroencephalogram, and Photoplethysmogram (PPG). Recent works, such as MOMENT, train a generalist time series foundation model with data from multiple domains, such as weather, traffic, and electricity. This paper aims to conduct a comprehensive benchmarking study to compare the performance of generalist and specialist models, with a focus on PPG signals. Through an extensive suite of total 51 tasks covering cardiac state assessment, laboratory value estimation, and cross-modal inference, we comprehensively evaluate both models across seven dimensions, including win score, average performance, feature quality, tuning gain, performance variance, transferability, and scalability. These metrics jointly capture not only the models' capability but also their adaptability, robustness, and efficiency under different fine-tuning strategies, providing a holistic understanding of their strengths and limitations for diverse downstream scenarios. In a full-tuning scenario, we demonstrate that the specialist model achieves a 27% higher win score. Finally, we provide further analysis on generalization, fairness, attention visualizations, and the importance of training data choice.

cs.LG↗

arXiv2025

Can Large Language Models Adequately Perform Symbolic Reasoning Over Time Series?

Uncovering hidden symbolic laws from time series data, as an aspiration dating back to Kepler's discovery of planetary motion, remains a core challenge in scientific discovery and artificial intelligence. While Large Language Models show promise in structured reasoning tasks, their ability to infer interpretable, context-aligned symbolic structures from time series data is still underexplored. To systematically evaluate this capability, we introduce SymbolBench, a comprehensive benchmark designed to assess symbolic reasoning over real-world time series across three tasks: multivariate symbolic regression, Boolean network inference, and causal discovery. Unlike prior efforts limited to simple algebraic equations, SymbolBench spans a diverse set of symbolic forms with varying complexity. We further propose a unified framework that integrates LLMs with genetic programming to form a closed-loop symbolic reasoning system, where LLMs act both as predictors and evaluators. Our empirical results reveal key strengths and limitations of current models, highlighting the importance of combining domain knowledge, context alignment, and reasoning structure to improve LLMs in automated scientific discovery. https://github.com/nuuuh/SymbolBench.

cs.AI↗