Search arXiv⌕ Search

arXiv subjects

Yang Liu

Publications and source records attributed to Yang Liu.

At least 73 records · Page 4Linked to original sources
arXiv2026

ChemArgus: Automated Scientific Diagnostics for Lab-Grade Multimodal Chemistry Reasoning: A Frontier Benchmark

Chemistry evaluation has moved past multiple-choice probes toward open-ended expert problems, yet the benchmarks that define the current frontier still score answer endpoints alone. Such scoring conflates memorized facts with stepwise derivation, treats visual and textual inputs as interchangeable, and confounds long-horizon planning with stepwise execution. These blind spots prevent any diagnosis of where the reasoning chain breaks in a model. We present ChemArgus, a competition-level chemistry benchmark drawn from chemistry competition material, textbooks, and research papers, whose scoring replaces the answer endpoint with three measurement properties. Fine-grained scoring grades every subquestion point by point with partial credit, so a score lands at the level of the step rather than the outcome. Multimodal capability is priced on every problem containing a real image, paired with a controlled text description, so visual understanding is measured against text on identical tasks. Diagnostic attribution reads paired metrics that share one scoring unit and differ only in context construction, so a gap between two scores isolates a single cause, namely integration, propagation, or local reasoning. GPT-5.6-Sol scores 23.2% on unassisted whole-problem solving against 28.8% and 30.4% under the two sequential metrics, a gap measuring the integration burden. Through rubric partial credit and these cause-isolating score gaps, ChemArgus shows not only whether a model answers chemistry correctly but where its reasoning chain fails, so that every score it reports doubles as a diagnosis, a readout of the chain rather than a verdict, aiming to guide language models step by step on their path toward chemistry expertise, one corrected step at a time, each step priced by the rubric, point by point.

cs.LG↗
arXiv2026

PrismaCoder: Pushing the Boundaries of Open Multimodal Code Intelligence through Decomposed Rubric Synthesis

Multimodal code intelligence connects the symbolic world of programs with the visual world their execution creates. Interest in this capability is burgeoning, as models are increasingly asked to produce charts, webpages, animations, and semi-structured documents. Recently, proprietary models have pushed far ahead, producing interactive artifacts even from a text description or a screenshot alone. Yet their data and training remain closed, and open progress is held back by the quantity and quality of available data and by the pipelines and methods for producing it. Quality is the harder of the two, since a sample must be correct as a program, well built, and above all visually right once rendered. No existing signal covers all three. Execution often returns just a pass or fail, code similarity does not map to the visual output, and task-specific verifiers do not generalize. To push open progress further, we start from the data side with PrismaForge, a hierarchical data pipeline centered on a synthesized rubric that decomposes these factors into checkable criteria. The same rubric serves as both filter and reward: it curates the supervised corpus, then drives reinforcement learning on visual-coding tasks. Leveraging PrismaForge for data synthesis and filtering, we build PrismaCoder-2M, the most diverse open dataset for multimodal code intelligence and the largest to date. On the model side, we train PrismaCoder-9B and PrismaCoder-35B on this data and release both. They lead the open state of the art by a wide margin and rival leading proprietary systems. Our analyses further show that the proposed rubric-based pipeline improves multimodal coding performance across model backbones. We release our code and data at https://qiushisun.github.io/PrismaCoder-Home/.

cs.CL↗
arXiv2026

Illusion of Alignment: Detecting Hidden Disagreement in Collaborative Dialogue

Collaborative dialogue can end with apparent agreement while participants still differ on goals, assumptions, or execution plans, creating an \textbf{illusion of alignment (IoA)}. A real-user study across 18 meetings confirms that IoA arises routinely in human collaboration. Yet IoA poses a paradox: if participants were aware of such disagreements, they would already be explicit; if not, they cannot articulate them when asked, leaving IoA invisible to both participants and observers. In this work, we make IoA detectable by generating diagnostic multiple-choice questions whose divergent answers across participants provide direct behavioral evidence of hidden disagreement. We construct \textbf{IoA-Suite}, a dataset and evaluation protocol for detecting hidden disagreement, spanning five task types and six domains. We find that even the best model attains only 49.5\% F1, with the bottleneck traced to private context that the dialogue does not surface. We then train \textbf{IoA-Prober-8B} based on IoA-Suite, reaching 51.8\% F1 on IoA-Suite. Across the aforementioned 18 real meetings, it surfaces 2.89 hidden disagreements per meeting that participants confirm they had not voiced, transferring to live human dialogue. Further, in multi-agent collaboration, pairing IoA-Prober-8B with LLM agents improves downstream task performance on BigCodeBench-Hard and HiddenBench.

cs.AI↗
arXiv2026

Benchmarking Self-Supervised 3D Medical Vision Foundation Models and Transferability Estimation

Self-supervised pretraining has expanded the pool of 3D medical foundation models, but published evaluations use different datasets, endpoints and fine-tuning protocols. Models also differ in architecture, pretraining objective and data, while downstream targets range from voxel-level masks to patient-level diagnoses and outcomes. As a result, published findings provide little directly comparable evidence for choosing a model for a new task. Direct comparison of adaptation performance requires fine-tuning candidates on labelled target data, at substantial computational cost. We benchmark 23 candidates across the same 14 segmentation, classification and regression tasks spanning multiple anatomies and modalities. No candidate ranked among the top three across all segmentation tasks or across all classification and regression tasks; a convolutional encoder could outperform a transformer trained with the same objective on one task but underperform it on another. We benchmark transferability estimators against the fine-tuned rankings and find that their agreement varies across dense segmentation, classification and regression. InsightRank uses task-appropriate frozen-feature statistics and architecture-aware pairwise comparisons to rank candidates before fine-tuning. Across the 23-model pool, mean Kendall tau-b was 0.431 for segmentation and 0.352 for classification and regression. The segmentation Top 3 contained a task winner on four of nine tasks; the image-level Top 3 contained none on the five tasks, despite better full-ranking agreement than the matched conventional estimators. A three-model shortlist reduced recorded extraction-plus-training job time by 85.8% across seven segmentation tasks. Together, the benchmark provides a common reference for model selection and InsightRank offers a lower-cost way to prioritize fine-tuning, with shortlist performance that differs by target type.

cs.CV↗
arXiv2026

Enhanced weak-field sensing using detuned squeezing

Achieving measurement precision beyond the standard quantum limit (SQL) is a fundamental object in quantum metrology. In this manuscript, we propose a streamlined protocol for weak signal sensing in a trapped-ion spin-oscillator system that simultaneously exploits detuned motional squeezing and spin-oscillator entanglement. By combining these two quantum resources, the phase accumulation in this spin system is significantly accelerated, as a result achieving an exponential improvement in the sensitivity over unsqueezed cases. Remarkably, we show that the significant metrological gain provided by our protocol persists in noisy environments, with the sensitivity remaining below the SQL. In particular, our protocol retains a distinct advantage in the short interrogation time regime, thus demonstrating the practical potential in quantum-enhanced sensing.

quant-ph↗
arXiv2026

Instructions Are Evidence, Not Objectives: Measuring Outcome Alignment of Agents

Large language model agents can follow complex human instructions to solve real-world tasks. In practice, however, initial instructions are often incomplete. In our analysis of 100 user-agent trajectories from SWE-chat, 56\% of initial instructions omit requirements that later emerge as necessary, and 52\% of the useful follow-up requirements could have been inferred from the project. A user states requirements from their own understanding of the task, and these \textbf{stated requirements} often differ from the \textbf{true requirements} that completing the task imposes. An agent should therefore treat the instruction as evidence rather than as the objective, inferring the true requirements from the instruction and the environment and satisfying them even when the user is not yet aware of all of them. We call this ability \textbf{outcome alignment}. In this work, we construct \textbf{TiGanEval}, a benchmark of software engineering and research tasks. In every task the user genuinely lacks part of the requirements, each requirement stays discoverable from the code and runtime behaviour of the environment, and requirements the agent infers are separated from those the user adds after observing delivered work. These properties together make outcome alignment measurable. Evaluating eight frontier models including GPT-5.6-Sol, we find outcome alignment broadly missing: the models lose 53 to 67 points of success rate when moving from the full specification to the partial instruction, and consequence feedback across four submissions does not close the gap. Furthermore, a plan-mode harness cannot compensate for the missing ability of the model itself, which places outcome alignment among the abilities the next generation of models should be developed for.

cs.AI↗
arXiv2026

Preliminary Exploration of Large Models Based on Self-Supervised Learning for Health Monitoring of Power Electronic Converters

Current artificial intelligence (AI) methods for power electronics converters make limited use of the abundant unlabeled data generated during real-world operation. This underexplored resource holds significant potential for advancing model performance. To address this gap, this paper presents PE-TSFM, a time-series foundation model tailored for power module health monitoring. The model adopts the PatchTST architecture combined with self-supervised learning and is trained on more than 143 million timestamps, where over 98\% of this training data is unlabeled and collected directly from an operating traction converter. In degradation classification tasks, PE-TSFM achieves an accuracy of 87\%, considerably surpassing five conventional time-series AI models whose performance ranges from 27\% to 45\%. These results demonstrate the model's strong few-shot learning capability and generalization across different operating conditions. This work promotes a shift from narrowly focused models in power electronics toward unified, scalable foundation models that can unlock value from large-scale unlabeled operational data.

eess.SY↗
arXiv2026

Revisiting Generative vs. Dense Retrieval: A Static Value Prior Reverses the Gap

Generative retrieval (GR) has emerged as a strong candidate generator for sequential recommendation as language models and semantic-item identifiers have improved. However, GR is not a uniformly stronger replacement of dense retrieval (DR) in online retail candidate generation. We examine their differences on two Amazon Toys datasets under different metrics. GR leads on best relevant item value among users both methods cover, while DR leads on exact-target recall and relevant value across the ranked list. With deep analysis, we find that popularity contributes the most to GR's measured value lead. Adding the same static item value to DR in one reranking step reverses this lead and keeps recall above GR. Finally, on an industrial-scale internal e-commerce dataset with LLM-judged relevance, we show that a 14B SID-LLM still underperforms a $47\times$ smaller dense encoder.

cs.IR↗
arXiv2026

PRB-WM: Physics-Regularized Beta World Models for Predictive Link Adaptation on Real Time-Varying Underwater Acoustic Channels

Adaptive modulation and coding (AMC) improves the spectral efficiency of underwater acoustic (UWA) links, but its feedback is inevitably stale: sound propagates slowly, and by the time channel state information (CSI) arrives, the fast time-varying multipath channel has moved on. Predictive link adaptation therefore needs more than a point forecast: it needs a distribution over each candidate action's consequences. We propose PRB-WM, a physics-regularized Beta world model for predictive per-subcarrier link adaptation on real time-varying UWA channels. A persistence-anchored stochastic latent-space dynamics model learns how the channel evolves; a counterfactual outcome head evaluates, for every candidate modulation and coding scheme (MCS), a Beta-binomial distribution of bit errors on shared imagined futures; and a chance-constrained receding-horizon model predictive controller maximizes expected spectral efficiency subject to a bounded probability of violating the bit error rate (BER) target. To the best of our knowledge, PRB-WM is the first world model for UWA communications. In closed-loop replay of six measured sea-trial links from the public UWA Channel Library against a 24-method pool spanning the full methodological spectrum, PRB-WM attains a mean goodput 87% above the strongest baseline, ranks first on all six time-varying links at the longest lookahead, and is the only method whose realized frame risk decreases systematically with lookahead, meeting the BER-violation contract where no baseline does. Ablations confirm that rollout sampling dispersion is essential to the chance constraint and that the physics anchor is necessary for trainability.

cs.SD↗
arXiv2026

Approaching Irrational Quantization

The integer and fractional quantum Hall effects reveal that quantized physical responses can take integer and fractional values, respectively. This naturally raises the question of whether quantization can extend beyond rational numbers to an irrational value. Here we show that such a quantized response can emerge from the limiting behavior of a finite, experimentally realizable geometric network. We develop a method based on infinite geometric structures, in which a sequence of finite networks rapidly converges toward an infinite limit. Following this approach, we identify a planar network whose normalized resistance converges to the irrational value $\sqrt{3}$, and realize a finite approximation with high-precision geometric control. The measured normalized resistance is $1.735 \pm 0.001\, h/e^2$ for a fully occupied Landau level in the designed pattern. Our experimental results extend the concept of quantization beyond rational values and demonstrate that an irrational quantized response can be approached through a finite physical system.

cond-mat.mes-hall↗
arXiv2026

Learning with a Single Rollout via Pass@k Policy Optimization

Group Relative Policy Optimization (GRPO) typically requires multiple rollouts per prompt to estimate relative credit, incurring substantial sampling cost and exacerbating training asynchrony. This limitation is particularly severe for scientific reasoning agents, where each trajectory may involve long-horizon reasoning, multiple turns, and parallel sub-runs. We study single-rollout reinforcement learning, where only one rollout is collected per prompt, and find that naive training can stagnate or become severely unstable. We introduce Single-Rollout Pass@k Policy Optimization (SR-PPO), which combines Monte Carlo token-level advantage estimation with Pass@k gradient reweighting. An online critic without large-scale pretraining is used both as a prefix-dependent baseline for credit assignment and to estimate the Pass@k gradient weight. Using only one rollout per prompt, SR-PPO achieves stable learning and performance comparable to GRPO across reasoning and logic benchmarks, even under off-policy drift that degrades the performance of standard Pass@1 PPO. A key factor in this stability is Pass@k reweighting, which emphasizes harder prompts while downweighting high-success prompts, for which imperfect credit estimates tend to produce noisier policy updates. Through analyses of critic quality, gradient noise, and update behavior, we investigate how Pass@k reweighting stabilizes single-rollout learning and why residual critic errors may limit SR-PPO's ability to outperform GRPO.

cs.LG↗
arXiv2026

We Have Been Interpreting Diffusion Transformer Backbone Behaviors Wrong

During a network forward call, diffusion backbones naturally produce many intermediate activations. If we evaluate each activation against an identical downstream benchmarking protocol, we can plot an inverted-U-shaped curve as the depth profile: the activation's performance goes up first and then down as we go into deeper layers. This is common knowledge in the diffusion model community, which inspires many applications including the widely-adopted REPA. However, the common interpretation of this depth profile may be inaccurate and incomplete: (i) The cause of this curve has long been treated as signal granularity changing from coarse to fine, but there has been little evidence in DiT backbones. (ii) The community has been mainly studying this phenomenon in a static way, without investigating how it changes across different training setups. In this work, we show a different mechanism picture for the profile, where the prediction target (e.g., $ε$, $u$, $x_0$) sits at the center. The semantically meaningful signal is only an intermediate by-product for predicting the target, while the prediction target is the real goal for the backbone. Furthermore, the choice of prediction targets as part of the training setup can predictably affect the curve's shape. With this updated interpretation, many applications of this depth profile can be better grounded and predicted. As an example, we show how to predict the near-optimal placement of REPA based on the prediction target of the training setup. The code is at https://github.com/Darkbblue/dit-compare.

cs.CV↗
arXiv2026

PEARL: Process-Experience Aware Reinforcement Learning for Multi-Turn Agents

Terminal rewards make multi-turn agents trainable without step-level annotations, but group-relative advantages provide little guidance about which intermediate decisions to adjust. Failed trajectories also contain reasonable actions, while unfavorable decision patterns can recur as training proceeds. We introduce PEARL (Process-Experience Aware Reinforcement Learning), which improves local credit assignment without an additional critic. Given two continuations from the same history with different outcomes, PEARL locates their first divergence and builds a feature representation of each branch's local decision. It clusters these decision features in recent failure and success memories to identify recurring local patterns. Using this pattern memory, PEARL selectively strengthens negative advantages at lower-reward decisions, while leaving positive advantages and the base policy optimizer unchanged. On AppWorld, PEARL improves over ARPO on Normal and Challenge tasks at both 4B and 8B scales. On ALFWorld, it achieves 97.9% mean success at 4B, 2.7 points above the strongest baseline, while remaining competitive at 8B. Ablations of shaping location, success-bank filtering, and memory window support the value of selective, recurring-pattern-based updates.

cs.AI↗
arXiv2026

Predict in Chunks, Revise in Spans: History-conditioned Action Regeneration for Robotic Manipulation

Action chunking improves temporal coherence in visuomotor imitation learning by jointly predicting sequences of future actions. However, many existing policies regenerate entire action chunks even when errors are localized, potentially replacing otherwise valid predictions. We introduce Contextual Masked Action Policy (CMAP), a visuomotor policy that refines continuous action chunks through history-conditioned local regeneration. Specifically, CMAP introduces a history-conditioned action tape that combines executed commands with masked future positions to produce coarse action predictions and per-step uncertainty scores. An uncertainty-guided span regeneration mechanism then selects a high-uncertainty contiguous span and reconstructs it using the executed history and retained future predictions as context. By jointly reconstructing neighboring actions, CMAP models local temporal dependencies while preserving all predictions outside the selected span. Extensive experiments on nine simulated and four real-world manipulation tasks demonstrate that CMAP achieves superior overall performance over several competitive baselines. Ablation studies validate the complementary contributions of its key components, while held-out trajectory analysis shows that local regeneration reduces prediction errors within selected spans.

cs.RO↗
arXiv2026

PyCache Trap: The Inspection-Execution Gap in Agent Skill Scanners

Agent skills combine instructions with executable resources, giving third-party packages access to an agent's runtime. Existing skill scanners inspect documentation and visible source, but Python may execute a bundled bytecode cache with different behavior. We study this gap between inspection and execution through PyCache Trap, which pairs benign source with a substituted cache accepted by the loader and connects it to a task-relevant invocation. Scanner-guided rewriting changes the invocation wording while preserving the cache body, separating package admission from recognition of the concealed behavior. Across 100 skills and seven scanners, PyCache Trap achieves 94-100% attack success, with no semantic recognition of the cache-resident behavior. We propose execution-aware validation (EAV) to connect inspected instructions, scripts, imports, and runtime artifacts in a typed execution graph. EAV combines grounded behavioral analysis with trusted reproduction of compiled artifacts. It detects all 100 evaluated source-present cache substitutions and reaches 92.8% Recall at 10.0% FPR across five attack families and 200 benign skills. The results support checking the executable artifacts a runtime can select as part of skill admission, within the supported loaders and code-object normalization. The code is released at https://github.com/leo0481/PyCacheTrap.

cs.CR↗
arXiv2026

Diffusion Precipitation Generation with Medium-range Weather Forecasting

Accurate daily precipitation forecasts are important for weather-sensitive decisions. Although recent Weather Foundation Models (WFMs) have achieved significant progress, they are trained on reanalysis data, which exhibit regional and intensity-dependent biases. Since retraining WFMs on observation-based targets is costly, we study Dynamics-Precipitation Decoupling, a plug-and-play framework that uses a separately trained generator to map non-precipitation forecasts from a frozen weather model to observation-aligned daily precipitation. Its key challenges are to capture heterogeneous rainfall distributions and align daily accumulations with atmospheric signals that evolve across space and time. To address them, we train a diffusion generator conditional on atmospheric states and satellite-gauge estimates of precipitation. A multi-scale recurrent encoder summarizes atmospheric evolution, while cross-attention aligns atmospheric and precipitation features during denoising. Extensive evaluation of various backbones, including Pangu-Weather, GraphCast, and HRES, on global and regional forecasting shows improved daily precipitation accuracy and event detection, with better representation of localized accumulation patterns. Global ensemble evaluation shows lower Brier scores than ENS and training-from-scratch GenCast at all reported lead times, with lower CRPS in most comparisons. Remarkably, a heavy-rainfall case study using 2,288 stations in China further supports improved daily accumulation patterns and event detection against ground observations.

cs.LG↗
arXiv2026

PILOT: Guiding Long-Horizon World-Model Planning for Robot Manipulation with Visual Subgoals and Policy Priors

Latent world-model planning has shown promising results in goal-conditioned robotic manipulation, but remains challenging for long-horizon planning due to limited temporal guidance and inefficient action search. A distant visual goal provides little guidance for preparatory actions, while uninformed action sampling makes local optimization inefficient. We propose PILOT (Policy-Initialized Local Optimization with Temporal subgoals), which uses a video model to generate intermediate visual subgoals and a learned policy to initialize the cross-entropy method (CEM). CEM refines candidate action sequences toward each subgoal using an action-conditioned world model in a receding-horizon control loop. In our primary real-robot evaluation across four UR5e manipulation tasks, PILOT achieves 68.8% success, compared with 55.0% without policy initialization, 13.8% for policy-only execution, and 0.0% for direct final-goal planning. Simulation results show that these gains persist over longer horizons: PILOT exceeds direct execution of the same policy by 30 percentage points on long Push-T tasks and by 22 percentage points on LIBERO-10. Policy initialization also accelerates local optimization: on Push-T, CEM matches the cost reached by generic initialization after 10 iterations in only 4 iterations and reduces the final planning cost by 24.1% under the same sampling budget. These results show that combining temporal subgoals with policy-guided action search improves both long-horizon task execution and local planning efficiency. Videos are available on the webpage: https://pilot-planner.github.io/.

cs.RO↗
arXiv2026

DISCERN: Identifiable Disease Evidence for Generalizable Medical Multimodal Learning

Clinical multimodal models can learn enrollment shortcuts shared across modalities and fail on new cohorts. We model these data through a latent structural causal model: disease state and background factors generate physiology through shared mechanisms, while cohorts differ in selection and measurement. Holding acquisition fixed, we identify conditional disease evidence without observing disease-state labels or recovering the full causal model. In common Gaussian coordinates, cross-cohort state alignment and natural-parameter changes identify a minimal selection statistic; conditioning on it removes enrollment-dependent evidence and yields an identifiable reference posterior within the specified shift family. State-dependent interactions can carry evidence invisible to individual-modality posteriors, motivating joint, information-preserving representations. DISCERN combines an information-lattice encoder, a latent-state model, and an outcome head through joint training. Under reversed synthetic enrollment on known latent coordinates, conditioning raises state AUROC from 0.043 to 0.883 and from 0.265 to 0.825, while random conditioning does not. Exploratory cross-database mortality experiments between eICU and MIMIC-IV improve AUROC over matched empirical risk minimization by 0.036 and 0.035. Learned-coordinate diagnostics distinguish preservation of state information from identification of the coordinates required by the estimator.

cs.LG↗