arXiv subjects

Wei Xia

Publications and source records attributed to Wei Xia.

At least 37 records · Page 2Linked to original sources

arXiv2026

StaminaBench: Stress-Testing Coding Agents over 100 Interaction Turns

We introduce StaminaBench, a benchmark that measures the stamina of coding agents: how many consecutive interaction turns (change requests) they can handle before failing. Unlike the prevailing fraction-of-tasks-solved metric, this matches real vibe-coding where sessions run dozens or hundreds of turns. In StaminaBench, agents implement a REST API server and modify it across a tunable number of procedurally generated follow-up change requests - 100 in our experiments, resulting in codebases of up to 6,000 lines. Tests are generated fully programmatically without LLM involvement, ensuring reproducibility and reliability; change sequences are drawn from either a hardcoded or LLM-driven sampler, both constrained to a structured action space to ensure changes are valid. The agent and the server run in an isolated environment and communicate with the benchmark through HTTP, making testing fully black-box and language-agnostic. We evaluate six agent harnesses paired with seven open-source LLMs across 20 scenarios of 100 turns each and find that: (1) all the tested models fail within 5-6 turns, confirming that vibe-coding-style programming without thorough testing produces bugs; (2) passing test feedback back to the agent and allowing it to retry improves passed turn count by up to 12x; and (3) a good harness is required for strong performance: stronger models exhibit up to a 6x gap between their best and worst harness, while weaker models fail with any harness. We release the benchmark and the generated tasks to enable further research into multi-turn coding agent behavior. Benchmark code and data: github.com/amazon-science/StaminaBench.

cs.SE↗

arXiv2026

When Do LLMs Reason? A Dynamical Systems View via Entropy Phase Transitions

Chain-of-thought (CoT) reasoning has become the default strategy for enhancing LLM capabilities, yet its application raises a fundamental question: when is explicit reasoning actually beneficial? Empirical evidence reveals a striking paradox: CoT often provides marginal or even negative gains on factual and open-ended tasks while multiplying token consumption. In this work, we show that LLM reasoning is not a static property of tasks or models, but a \emph{dynamic decoding state} that emerges during generation. Through systematic analysis, we find early-stage entropy dynamics provide a reliable signal of this state: tasks benefiting from CoT exhibit consistent entropy reduction, while others display unstable or increasing patterns. This behavior can be interpreted as a phase-transition-like shift from a high-entropy exploratory regime to a low-entropy structured reasoning regime. Based on these insights, we propose \textbf{EDRM} (Entropy Dynamics-based Reasoning Manifold), a lightweight and training-free routing framework that leverages early decoding entropy to adaptively select inference strategies. EDRM embeds entropy trajectories into a compact and interpretable manifold representation, enabling both zero-shot deployment and fine-grained instance-level adaptation. Across 15 benchmarks and 4 LLMs of varying scales and architectures, EDRM consistently outperforms static baselines. At the dataset level, EDRM achieves \textbf{41--55\%} token reduction while improving accuracy with as few as 50 calibration samples. At the instance level, it further improves accuracy by up to \textbf{4.7\%} while maintaining \textbf{27--45\%} token savings. These results suggest that reasoning should be invoked selectively rather than by default, and demonstrate the effectiveness of entropy-driven decoding control for efficient and adaptive LLM inference.

cs.LG↗

arXiv2026

Time series learning in a many-body Rydberg system with emergent collective amplification

Interacting Rydberg atoms constitute a versatile platform for the realization of non-equilibrium states of matter. Close to phase transitions, they respond collectively to external perturbations, which can be harnessed for technological applications in the domain of quantum metrology and sensing. Owing to the controllable complexity and straightforward interpretability of Rydberg atoms, we can observe and tune the emergent collective amplification. Here, we investigate the application of an interacting Rydberg vapour for the purpose of time series prediction. The vapour is driven by a laser field whose Rabi frequency is modulated in order to input the time series. We find that close to a non-equilibrium phase transition, where collective effects are amplified, the capability of the system to learn the input becomes enhanced. This is reflected in an increase of the accuracy with which future values of the time series can be predicted. Using the Lorenz time series and temperature data as examples, our work demonstrates how emergent phenomena enhance the capability of noisy many-body systems for data processing and forecasting.

quant-ph↗

arXiv2026

DLLG: Dynamic Logit-Level Gating of LLM Experts

Leveraging multiple specialized LLMs can combine complementary strengths, but existing approaches trade adaptability for stability: routing commits prematurely, heuristic ensembling depends on fragile proxies, and parameter merging introduces interference. We propose DLLG (Dynamic Logit-Level Gating), a dynamic logit-level ensembling framework that learns token-level expert fusion from sparse response-level supervision. A lightweight gating module predicts step-wise fusion weights, linking trajectory-level correctness to generation without token-level labels or expert retraining. Across diverse reasoning and code benchmarks, DLLG consistently outperforms strong routing, heuristic ensembling, and parameter-merging baselines across model scales, highlighting learned logit-level fusion as a robust and scalable paradigm for integrating specialized experts.

cs.CL↗

arXiv2026

Cesarean Scar Defect Segmentation in Transvaginal Ultrasound Images: a Dataset and Benchmark

Cesarean Scar Defect (CSD) is one of the most prevalent complications following cesarean delivery. Transvaginal ultrasonography is widely used for primary CSD screening. Accurate determination of CSD outline and dimensions is crucial for treatment. However, CSDs are frequently overlooked by sonographers due to small size and irregular morphology, suboptimal image quality, and limited clinical awareness in resource-constrained settings. Despite artificial intelligence advances in medical imaging, no public dataset exists for transvaginal ultrasound CSD segmentation. To address this gap, we present a comprehensive CSD dataset comprising 1,111 images and 16 videos, yielding 501 positive samples with confirmed CSD and precise pixel-level manual annotations. Annotations are performed following standardized clinical guidelines through collaboration between experienced sonographers and trained PhD students. This work provides high-quality benchmark resources for advancing medical image segmentation algorithms and promoting clinical innovation. Ultimately, improved CSD diagnosis and subsequent treatment strategies can enhance the quality of life in women of reproductive age, representing significant value for both medical research and clinical practice.

cs.CV↗

arXiv2026

Meta-Cognitive Memory Policy Optimization for Long-Horizon LLM Agents

Memory-augmented LLM agents tackle complex long-horizon tasks by recursively summarizing interaction trajectories into compact memory. However, existing approaches typically train these memory policies using outcome-based reinforcement learning, failing to localize where intermediate memory quality degrades. As interactions unfold, ambiguous recursive summaries progressively discard task-relevant information and introduce semantic noise. This exacerbates belief deviation, obscuring the agent's estimate of the latent task state and ultimately derailing long-horizon reasoning. We therefore argue that memory optimization should focus not merely on trajectory-level success, but on the clarity of the belief induced by intermediate summaries. To this end, we introduce Belief Entropy, a self-supervised proxy that probes how uncertain the model remains about the latent task state given its current memory. Based on this proxy, we propose Metacognitive Memory Policy Optimization (MMPO). Instead of relying only on sparse outcome-based signals, MMPO provides fine-grained, memory-specific supervision via explicitly penalizing summaries that induce high epistemic uncertainty. Experiments show that MMPO consistently outperforms existing methods on diverse long-horizon tasks, maintaining 97.1% performance even when scaled to 1.75M-token contexts.

cs.AI↗

arXiv2026

EvoMAS: Evolutionary Generation of Multi-Agent Systems

Large language model (LLM)-based multi-agent systems (MAS) show strong promise for complex reasoning, planning, and tool-augmented tasks, but designing effective MAS architectures remains labor-intensive, brittle, and hard to generalize. Existing automatic MAS generation methods either rely on code generation, which often leads to executability and robustness failures, or impose rigid architectural templates that limit expressiveness and adaptability. We propose Evolutionary Generation of Multi-Agent Systems (EvoMAS), which formulates MAS generation as structured configuration generation. EvoMAS performs evolutionary generation in configuration space. Specifically, EvoMAS selects initial configurations from a pool, applies feedback-conditioned mutation and crossover guided by execution traces, and iteratively refines both the candidate pool and an experience memory. We evaluate EvoMAS on diverse benchmarks, including BBEH, SWE-Bench, and WorkBench, covering reasoning, software engineering, and tool-use tasks. EvoMAS consistently improves task performance over both human-designed MAS and prior automatic MAS generation methods, while producing generated systems with higher executability and runtime robustness. EvoMAS outperforms the agent evolution method EvoAgent by +10.5 points on BBEH reasoning and +7.1 points on WorkBench. With Claude-4.5-Sonnet, EvoMAS also reaches 79.1% on SWE-Bench-Verified, matching the top of the leaderboard. Code is available at https://github.com/amazon-science/EvoMAS

cs.LG↗

arXiv2026

Gated KalmaNet: A Fading Memory Layer Through Test-Time Ridge Regression

Linear State-Space Models (SSMs) offer an efficient alternative to softmax Attention with constant memory and linear compute, but their lossy, fading summary of the past hurts recall-oriented tasks. We propose Gated KalmaNet (GKA, pronounced "gee-ka"), a layer that accounts for the full past while retaining SSM-style efficiency. We ground our approach in the Kalman Filter (KF), and show that several existing SSM layers (DeltaNet, Gated DeltaNet, Kimi Delta Attention) are approximations to the KF recurrence under an identity error covariance assumption, which ignores how past keys and values should optimally influence state updates. In contrast, GKA maintains the full error covariance and computes the exact Kalman gain. Under a steady-state assumption that enables parallelization, this reduces to an online ridge regression with constant memory and linear compute. The standard KF equations are numerically unstable in low-precision settings (e.g., bfloat16) and hard to parallelize on GPUs. We address this with (1) adaptive regularization via input-dependent gating to control the ridge regression's condition number, and (2) Chebyshev Iteration, which we show is more stable than conventional iterative solvers in low precision. We further develop hardware-aware chunk-wise kernels for efficient training. Empirically, GKA outperforms existing SSM layers (e.g., Mamba2, Gated DeltaNet) on short-context tasks and achieves more than 10\% relative improvement on long-context RAG and LongQA up to 128k tokens. We further show GKA outperforms Mamba when extended to ImageNet classification. Our code, including Triton kernels for training and inference (vLLM), along with a model zoo of GKA-based Hybrid models at 8B and 32B scale on HuggingFace, is released under Apache 2.0.

cs.LG↗

arXiv2026

Experience Sharing in Mutual Reinforcement Learning for Heterogeneous Language Models

We introduce Mutual Reinforcement Learning, a framework for concurrent RL post-training in which heterogeneous LLM policies exchange typed experience while keeping separate parameters, objectives, and tokenizers. The framework combines a Shared Experience Exchange (SEE), Multi-Worker Resource Allocation (MWRA), and a Tokenizer Heterogeneity Layer (THL) that retokenizes text and aligns token-level traces across incompatible vocabularies. This substrate makes the experience-sharing design question operational across model families. We instantiate three controlled probes on top of GRPO: data-level rollout sharing via Peer Rollout Pooling (PRP), value-level advantage sharing via Cross-Policy GRPO Advantage Sharing (XGRPO), and outcome-level success transfer via Success-Gated Transfer (SGT). A contextual-bandit analysis characterizes their structural positions on a stability-support trade-off: PRP pays density-ratio variance and THL residual costs, XGRPO preserves on-policy actor support while changing scalar baselines, and SGT supplies a rescue-set score direction toward verified peer successes. In the evaluated regime, outcome-level sharing occupies the favorable point of this trade-off.

cs.LG↗

arXiv2026

Priming: Hybrid State Space Models From Pre-trained Transformers

Hybrid State-Space models combine Attention with recurrent State-Space Model (SSM) layers, balancing eidetic memory from Attention with compressed fading memory from SSMs. This yields smaller Key-Value caches and faster decoding than Transformers, along with a richer architectural design space. Exploring that design space at scale has so far required training from scratch, a barrier that has kept most large-model Hybrid research within a narrow range of architectures. We introduce Priming, a method that turns Hybrid architecture design from a pre-training problem into a knowledge transfer one. Priming initializes a Hybrid model from a pre-trained Transformer and, through short alignment and post-training phases, recovers downstream quality using less than 0.5% of the source model's pre-training token budget. Priming is agnostic to the source Transformer family (e.g., Qwen, Llama, Mistral), model class (dense or Mixture-of-Experts), and model scale. Priming enables us to run the first controlled comparison of SSM layer types at scale under identical conditions. We evaluate, Gated KalmaNet (GKA), Gated DeltaNet (GDN), and Mamba-2, and show that their expressiveness hierarchy, GKA>GDN>Mamba-2, directly predicts downstream performance on long-context reasoning tasks. We scale Priming to 8B/32B reasoning models with native 128K contexts. Our Hybrid GKA 32B improves over its source Qwen3-32B by +3.8 average reasoning points, while staying within 1% of a Transformer post-trained on the same data and enabling up to 2.3x higher decode throughput. To foster research on Hybrid architectures, we release a model zoo of primed Hybrid models for long-context reasoning and instruction following, together with the Priming training and inference code (Sequence Parallelism algorithms for long-context training, optimized GKA kernels, and vLLM serving plugin), all under Apache~2.0 License.

cs.LG↗

arXiv2026

Breaking the Trade-off: Bulk 2D Ising Superconductivity with High Tc and Giant Interlayer Spacing via a Unique Chain Intercalation in (BaS)1/3TaS2

Two-dimensional (2D) transition metal dichalcogenides (TMDs) are promising platforms for low dimensional superconductivity. However, in conventional intercalated systems, achieving a high superconducting transition temperature (Tc) often comes at the expense of reduced interlayer spacing and weakened 2D character. Here, we overcome this long-standing compromise through a unique chain-like intercalation strategy. We report the synthesis and properties of a new polymorph, (BaS)1/3TaS2, in which a distinctive Ba-S-S-Ba chain structure is inserted between TaS2 bilayers. This unique configuration breaks the bulk c axis mirror symmetry while achieving exceptional interlayer decoupling, with an inter-bilayer spacing of 12.75 Å-more than three times that of pristine 2H-TaS2. By suppressing interlayer electronic coupling, this structural evolution allows local inversion symmetry breaking within individual TaS2 layers to dominate. This prevents compensation of the Ising spin-orbit fields typical of centrosymmetric bulk phases, enabling robust 2D Ising superconductivity. Remarkably, the compound exhibits an enhanced Tc without sacrificing its large interlayer spacing, thereby breaking the conventional trade-off between large spacing/high anisotropy and high Tc. Comprehensive transport, magnetic, and thermodynamic measurements confirm its robust superconducting state. Our work establishes a versatile intercalation framework for designing bulk-like 2D Ising superconductors, providing a new route to reconcile competing material demands and expanding the scope of Ising superconductivity research.

cond-mat.supr-con↗

arXiv2026

TeamLLM: A Human-Like Team-Oriented Collaboration Framework for Multi-Step Contextualized Tasks

Recently, multi-Large Language Model (LLM) frameworks have been proposed to solve contextualized tasks. However, these frameworks do not explicitly emulate human team role division, which may lead to a single perspective, thereby weakening performance on multi-step contextualized tasks. To address this issue, we propose TeamLLM, a human-like Team-Oriented Multi-LLM Collaboration Framework. TeamLLM adopts four team roles with distinct division and employs a three-phase multi-LLM collaboration for multi-step contextualized tasks. To evaluate the effectiveness of TeamLLM on multi-step contextualized tasks, we propose Contextually-Grounded and Procedurally-Structured tasks (CGPST) and construct the CGPST benchmark. This benchmark has four core features: contextual grounding, procedural structure, process-oriented evaluation and multi-dimensional assessment. We evaluate ten popular LLMs on CGPST at overall-level, step-level, and dimension-level. Results show that TeamLLM substantially improves performance on CGPST. We release the benchmark with scenarios, full-process responses and human scores from ten LLMs. The code and data are available at https://anonymous.4open.science/r/TeamLLM-anonymous-C50E/.

cs.CL↗

arXiv2026

Asymmetric Actor-Critic for Multi-turn LLM Agents

Large language models (LLMs) exhibit strong reasoning and conversational abilities, but ensuring reliable behavior in multi-turn interactions remains challenging. In many real-world applications, agents must succeed in one-shot settings where retries are impossible. Existing approaches either rely on reflection or post-hoc evaluation, which require additional attempts, or assume fully trainable models that cannot leverage proprietary LLMs. We propose an asymmetric actor-critic framework for reliable conversational agents. A powerful proprietary LLM acts as the actor, while a smaller open-source critic provides runtime supervision, monitoring the actor's actions and intervening within the same interaction trajectory. Unlike training-based actor-critic methods, our framework supervises a fixed actor operating in open-ended conversational environments. The design leverages a generation-verification asymmetry: while high-quality generation requires large models, effective oversight can often be achieved by smaller ones. We further introduce a data generation pipeline that produces supervision signals for critic fine-tuning without modifying the actor. Experiments on $τ$-bench and UserBench show that our approach significantly improves reliability and task success over strong single-agent baselines. Moreover, lightweight open-source critics rival or surpass larger proprietary models in the critic role, and critic fine-tuning yields additional gains over several state-of-the-art methods.

cs.CL↗

arXiv2026

High-Accuracy Temporal Prediction via Experimental Quantum Reservoir Computing in Correlated Spins

Physical reservoir computing provides a powerful machine learning paradigm that exploits nonlinear physical dynamics for efficient information processing. By incorporating quantum effects, quantum reservoir computing offers superior potential for machine learning applications, as quantum dynamics are exponentially costly to simulate classically. Here, we present a novel quantum reservoir computing approach based on correlated quantum spin systems, exploiting natural quantum many-body interactions to generate reservoir dynamics, thereby circumventing the practical challenges of deep quantum circuits. Our experimental implementation supports nontrivial quantum entanglement and exhibits sufficient dynamical complexity for high-performance machine learning. We achieve state-of-the-art performance in experiments on standard time-series benchmarks, reducing prediction error by 1 to 2 orders of magnitude compared to previous quantum reservoir experiments. In long-term weather forecasting, our 9-spin quantum reservoir delivers greater prediction accuracy than classical reservoirs with thousands of nodes. This represents the first experimental demonstration of quantum machine learning outperforming large-scale classical models on real-world tasks.

quant-ph↗

arXiv2026

Ostrakon-VL: Towards Domain-Expert MLLM for Food-Service and Retail Stores

Multimodal Large Language Models (MLLMs) have recently achieved substantial progress in general-purpose perception and reasoning. Nevertheless, their deployment in Food-Service and Retail Stores (FSRS) scenarios encounters two major obstacles: (i) real-world FSRS data, collected from heterogeneous acquisition devices, are highly noisy and lack auditable, closed-loop data curation, which impedes the construction of high-quality, controllable, and reproducible training corpora; and (ii) existing evaluation protocols do not offer a unified, fine-grained and standardized benchmark spanning single-image, multi-image, and video inputs, making it challenging to objectively gauge model robustness. To address these challenges, we first develop Ostrakon-VL, an FSRS-oriented MLLM based on Qwen3-VL-8B. Second, we introduce ShopBench, the first public benchmark for FSRS. Third, we propose QUAD (Quality-aware Unbiased Automated Data-curation), a multi-stage multimodal instruction data curation pipeline. Leveraging a multi-stage training strategy, Ostrakon-VL achieves an average score of 60.1 on ShopBench, establishing a new state of the art among open-source MLLMs with comparable parameter scales and diverse architectures. Notably, it surpasses the substantially larger Qwen3-VL-235B-A22B (59.4) by +0.7, and exceeds the same-scale Qwen3-VL-8B (55.3) by +4.8, demonstrating significantly improved parameter efficiency. These results indicate that Ostrakon-VL delivers more robust and reliable FSRS-centric perception and decision-making capabilities. To facilitate reproducible research, we will publicly release Ostrakon-VL and the ShopBench benchmark.

cs.AI↗

arXiv2026

Multi-Constrained Evolutionary Molecular Design Framework: An Interpretable Drug Design Method Combining Rule-Based Evolution and Molecular Crossover

This study proposes MCEMOL (Multi-Constrained Evolutionary Molecular Design Framework), a molecular optimization approach integrating rule-based evolution with molecular crossover. MCEMOL employs dual-layer evolution: optimizing transformation rules at rule level while applying crossover and mutation to molecular structures. Unlike deep learning methods requiring large datasets and extensive training, our algorithm evolves efficiently from minimal starting molecules with low computational overhead. The framework incorporates message-passing neural networks and comprehensive chemical constraints, ensuring efficient and interpretable molecular design. Experimental results demonstrate that MCEMOL provides transparent design pathways through its evolutionary mechanism while generating valid, diverse, target-compliant molecules. The framework achieves 100% molecular validity with high structural diversity and excellent drug-likeness compliance, showing strong performance in symmetry constraints, pharmacophore optimization, and stereochemical integrity. Unlike black-box methods, MCEMOL delivers dual value: interpretable transformation rules researchers can understand and trust, alongside high-quality molecular libraries for practical applications. This establishes a paradigm where interpretable AI-driven drug design and effective molecular generation are achieved simultaneously, bridging the gap between computational innovation and practical drug discovery needs.

cs.NE↗

arXiv2026

AP2O-Coder: Adaptively Progressive Preference Optimization for Reducing Compilation and Runtime Errors in LLM-Generated Code

LLMs' code generation capabilities have yielded substantial improvements in the effectiveness of programming tasks. However, LLM-generated code still suffers from compilation and runtime errors. Existing offline preference optimization methods primarily focus on enhancing LLMs' coding abilities using pass/fail signals in the preference data, overlooking the deep-level error types in the failed codes. To address this, we propose Adaptively Progressive Preference Optimization (AP2O) for coding (i.e., AP2O-Coder), a method that guides LLMs adaptively and methodically to reduce code errors for code generation. Specifically, we construct an error notebook from failed codes and progressively optimize the LLM to correct errors type by type. Furthermore, we adaptively replay error types to tailor to the LLM's changing weaknesses throughout the training process. Through extensive experiments on both code and general LLMs (Llama, Qwen, and DeepSeek series) with parameters ranging from 0.5B to 34B, our AP2O-Coder improves code generation performance by up to 3% in pass@k while using less preference data. Code: https://github.com/TsingZ0/AP2O

cs.SE↗

arXiv2026

GAPO: Robust Advantage Estimation for Real-World Code LLMs

Reinforcement learning (RL) is widely used for post-training large language models (LLMs) in code editing, where group-relative methods, such as GRPO, are popular due to their critic-free and normalized advantage estimation. However, in real-world code-editing scenarios, reward distributions are often skewed with unpredictable noise, leading to distorted advantage computation and increased rollout outliers. To address this issue, we propose Group Adaptive Policy Optimization (GAPO), which adaptively finds an interval with the highest SNR (Signal to Noise Ratio) per prompt and uses the median of that interval as an adaptive Q to replace the group mean in advantage calculation to reduce noise further. This adaptive Q robustly handles rollout noise while remaining plug-and-play and efficient. We evaluate GAPO on nine instruction-tuned LLMs (3B-14B) using a collected large dataset of 51,844 real-world, history-aware code-editing tasks spanning 10 programming languages. GAPO yields up to 4.35 in-domain (ID) and 5.30 out-of-domain (OOD) exact-match improvements over GRPO and its variant DAPO, while achieving lower clipping ratios and higher GPU throughput. Code: https://github.com/TsingZ0/verl-GAPO.

cs.LG↗