Searcharxiv⌕ Search

arXiv subjects

Zhiyuan Wang

Publications and source records attributed to Zhiyuan Wang.

At least 19 recordsLinked to original sources

Stability-Compensated Threshold Calibration for Selection-Conditioned Risk Control in Large Language Models

Large language models often produce hallucinated answers, motivating selective prediction policies that abstain on uncertain queries. However, heuristic uncertainty thresholds alone do not provide statistical control of errors among accepted predictions. Linear Expectation Constraints (LEC) reduce selection-conditioned error rate (SCER) calibration to a nonpositive expected linear-loss constraint, but the associated loss is non-monotonic in the acceptance threshold. We construct an exact counterexample under i.i.d. sampling demonstrating that LEC's empirical threshold-selection rule can violate the population SCER target. To address this risk-transfer gap, we introduce Stability-Compensated Linear Expectation Control (SC-LEC), which uses the negative linear risk of a theoretical full-data reference to compensate for threshold-selection instability. For local expansion beyond a safe anchor, we derive two finite-sample certificates: a stability-based certificate exploiting rank crossings and a direct certificate using the structure of threshold expansion to reduce conservatism. With a valid population anchor margin, our certificate conditions ensure finite-sample marginal SCER control, and independent margin certification further provides a $1-δ$ high-probability guarantee for this calibration-averaged control. Experiments on CommonsenseQA, ARC-Challenge, and TriviaQA show useful retention gains beyond the same certified anchor, while revealing conservative behavior in small-margin regimes.

cs.CL↗

Constructing Generalizable Parallel Algorithm Portfolios for Multi-Objective Binary Optimization via General-Purpose Instance Generation

Despite progress in constructing generalizable parallel algorithm portfolios~(PAPs), no general-purpose method is yet available for multi-objective binary optimization problems~(MOBOPs) without problem-specific customization. To fill this gap, this work proposes domain-agnostic co-evolution of parameterized search for multi-objective binary optimization~(DACMO), the first general-purpose method for automatic PAP construction on MOBOPs. Its core, Unified Instance Representation~(UIR), represents and generates MOBOP instances with varying dimensions without knowledge of problem-specific instance structures. UIR separates common features of the target class from instance-specific features, preserving the common features learned from real instances during generation. DACMO co-evolves the training instances and PAP by alternating UIR-based instance augmentation with PAP updates. Experiments cover four structurally different MOBOP classes, including two black-box classes from real-world applications. Starting from five real training instances per class, DACMO constructs high-performance PAPs without problem-specific customization. On unseen instances from all four classes, these PAPs outperform portfolios of tuned classical multi-objective evolutionary algorithms and those constructed by the recent automatic PAP construction method EoH-S. DACMO also achieves overall performance comparable to the strong CEPS baseline and significantly outperforms it on two classes, without the substantial expert effort CEPS requires for problem-specific instance generator design.

cs.NE↗

KVPath: Connecting Key-Value Representations Across Transformer Layers

Transformers typically cache an independent key-value (KV) representation for each token at every layer. Recent work reveals that a token's KV representations across neighboring layers are correlated and contain shared information. We hypothesize that this correlation implies incremental inter-layer changes, which can be effectively captured by low-dimensional deltas. Motivated by this, we propose KVPath, a new KV representation paradigm that connects KV representations across the network depth into a path. The path establishes a full-dimensional base KV at the first layer, while subsequent layers generate low-dimensional delta KVs to update the preceding KV, thereby constructing their own KV states. By caching the base KV once alongside these compact delta KVs, KVPath preserves distinct layer-specific representations without storing full KV states at every layer. We develop two variants of KVPath: KVPath-Residual, which updates vector values via residual addition, and KVPath-Rolling, which progressively updates feature dimensions in a rolling manner. Under matched parameter counts, KVPath reduces KV-cache storage by up to 68\% for Grouped-Query Attention (GQA) and 55\% for Multi-Head Latent Attention (MLA), while maintaining or even improving model performance.

cs.LG↗

Trapping of nanofabricated SiO$_2$ high-aspect-ratio discs in an optical lattice

We describe the nanofabrication and loading of lithographically patterned, high aspect ratio $10$ $μ$m diameter silicon dioxide discs with thickness of $400$ nm into an optical standing wave trap in vacuum. We obtain motional spectra of the center of mass motion of the discs using a combination of interferometric and imaging methods. The method is capable of high-efficiency loading and optical trapping using small numbers of objects, of the order of a few particles, with customized geometries and sizes compatible with wafer scale nanofabrication methods. Disc-like particles in optical traps have been identified as promising candidates for detectors of high frequency gravitational waves or as sensitive accelerometers.

physics.optics↗

AMSchem: Automated Schematic Visualization of AMS Circuits via Verifiable Agentic Generation

For decades, the netlist-to-schematic gap has remained a persistent challenge in design automation, especially for analog and mixed-signal (AMS) circuits. Netlists encode connectivity, not visual presentation, making their manual conversion into readable schematics time-consuming and dependent on expert conventions. The same gap leaves LLM-based design tools without structured visual feedback, forcing them to iterate on circuits they cannot see. We present AMSchem, a training-free, verifier-guided agentic framework that automatically generates human-readable schematics from AMS netlists. AMSchem separates circuit-aware visual planning (DrawingIntent) from executable schematic geometry (SchematicIR), and applies localized visual feedback with bounded, revalidated repair patches. Deterministic validators enforce device preservation, pin-net consistency, and geometric correctness throughout. For evaluation, we construct AMSchem-Bench comprising 64 circuits of 12 types. On the 40-circuit schematic subset, AMSchem with gpt-5.5 achieves 85.0\% validity (vs. 32.5\% direct generation). When integrated as a visual feedback module in LLM-based analog design, AMSchem resolves all 13 previously failed cases from AnalogCoder benchmark across five models, substantially improving the success rate. Our dataset and benchmark are available at https://www.amschem.com/benchmark

cs.AI↗

Cooperative Coevolution for Resource-Constrained Agentic LLM Post-Training

Tool-using large language model (LLM) agents produce long, multi-turn trajectories, making gradient-based post-training memory-intensive. Evolution strategies (ES) enable memory-efficient full-parameter post-training without backpropagation and can eventually match the performance of gradient-based reinforcement learning (RL). However, resource-constrained settings typically offer only a few GPUs, so the high GPU-hour requirements of ES translate into prohibitively long training times. To address this, we introduce Cooperative Parameter-subspace Evolution Strategy (CoPES), a cooperative coevolutionary method that decomposes the full parameter space into lower-dimensional subspaces and searches over them cooperatively to improve optimization efficiency. On Qwen3.5-4B math, CoPES recovers 92% of full-parameter GRPO's validation-accuracy gain within the GPU-hour budget of GRPO's best validation checkpoint, versus 67% for standard ES. Its theoretical GPU memory requirement is less than one-eighth that of full-parameter GRPO. It consistently outperforms standard ES and LoRA-based GRPO on all evaluated pass@k metrics across five math benchmarks. CoPES's gains extend to Qwen3.5-9B with unchanged optimization hyperparameters and to multi-hop QA. These results demonstrate an improved trade-off between memory requirements and training time for agentic LLM post-training under resource constraints. The code is open-sourced in https://github.com/MetaronWang/CoPES

cs.AI↗

Momentum-space engineering in van der Waals heterostructures

Twenty years of graphene research have established Dirac fermions as a versatile platform for exploring quantum topological phenomena. Yet control over graphene's low-energy spectrum mostly follows the route of miniband engineering using real-space superlattices. Here we introduce a complementary approach based on momentum-space 'Umklapp' band engineering. This is realised by matching reciprocal space positions of band edges of two adjacent highly incommensurate 2D crystals to coincide upon a boost from their Bragg vectors, creating resonant interlayer hybridisation that qualitatively reconstructs their spectra. To demonstrate this approach experimentally, we use monolayer graphene on MoS$_2$, twisted such that MoS$_2$ Q-valley states are brought by Umklapp into energy-momentum resonance with graphene's Dirac cones. The ensuing hybridization results in the formation of a clear additional Dirac-like spectral feature, identified by a distinct Landau-level fan and Shubnikov-de Haas oscillations characteristic to Berry phase $π$ electrons. These new topological spectral features create a spin-valley locked Ising system which is robust against small twist-angle variations that can be further tuned by an external displacement field. More broadly, Umklapp band engineering offers a practical strategy for tailoring the electronic structure in van der Waals heterostructures, expanding the design space of quantum materials beyond moiré superlattices and weak proximity effects.

cond-mat.mes-hall↗

Where Prefill Expert Reduction Hurts MoE Models: A Few Prompt Tokens Carry Most of the Short-Answer F1 Loss

Running prompt tokens through fewer routed experts during prefill makes mixture-of-experts (MoE) serving cheaper, but what it damages has not been measured. We locate the damage with token-, layer- and attention-path interventions in four MoE families on short-answer question answering. At half the experts the F1 loss is mostly answer form: the model still finds the answer but no longer answers briefly. The loss is made at a few tokens: the two chat-template tokens before the answer in Qwen3-30B-A3B, and the first 8-32 tokens, where the instruction sits, in DeepSeek-V2-Lite. Keeping them at full experts removes most of the F1 loss and about half and a fifth of the increase in answer NLL. Reduction changes the first-token sink statistics by less than 4%, and restoring super experts or massive activations recovers at most a quarter of the loss. We then test placement on real prompts with about 4% of tokens at full experts. Putting them on the task interface keeps F1 within one point of the unmodified model, while the same budget scattered in the documents is about 3 points worse in both prespecified primary tests of this comparison. Time to first token falls about 14% on 11K-token prompts.

cs.CL↗

GlucoAgent: Personalized Glucose Forecasting from Multimodal Personal Context

Continuous glucose monitors, smartwatches, meal logs, and health profiles form a rich multimodal personal context that can significantly enhance glucose forecasting. However, existing methods struggle to effectively utilize these signals to improve glucose prediction because the relevance of different modalities varies drastically across individuals and time. We present GlucoAgent, an LLM-based agentic system that addresses this challenge through dynamic, case-by-case analysis at inference time. For each forecast, GlucoAgent writes and runs code over an individual's multimodal context and returns predictions with a reason and supporting evidence. The generated programs are highly diverse, ranging from retrieving similar past meals to fitting and back-testing physiological and statistical models. On CGMacros, a multimodal glucose dataset, GlucoAgent outperforms time series foundation models on every forecasting horizon, achieving an average RMSE of 22.5 mg/dL compared to 23.5 mg/dL of Chronos-2, the leading multimodal time series foundation model. Furthermore, GlucoAgent is highly effective at leveraging multimodal data. The addition of multimodal data reduces its average RMSE by 1.4 mg/dL (24.0 mg/dL to 22.5 mg/dL), compared to a marginal 0.3 mg/dL (23.8 mg/dL to 23.5 mg/dL) improvement for Chronos-2. This advantage is most pronounced in challenging post-meal windows, where GlucoAgent uses multimodal context to reduce 60-minute forecast RMSE by 4.6 mg/dL (31.0 mg/dL to 26.5 mg/dL), far exceeding the 0.8 mg/dL (31.9 mg/dL to 31.1 mg/dL) reduction achieved by Chronos-2.

cs.AI↗

Dual Dimensionality for Local and Global Attention

Decoder-only Transformers compute attention over the KV cache of preceding tokens. Keys and Values are typically represented with the same dimensionality, regardless of their distance from the prediction target. In natural language, however, the next word is most strongly influenced by the immediately preceding tokens. We hypothesize that local and distant tokens impose asymmetric demands on representational capacity: local tokens are more critical for predicting immediate outputs and thus require richer representations, whereas distant tokens primarily serve as long-range memory, for which lower-dimensional representations may suffice. We formalize this idea as Distance-Adaptive Representation (DAR), which preserves full-dimensional representations within a local context window while assigning reduced-dimensional representations (e.g. 1/2 or 1/4 of the original dimensionality) to tokens beyond that window. Across multiple pretraining scales (70M to 410M parameters), as well as continued supervised fine-tuning on a 1B-scale model, this approach closely matches the performance of full-dimensional baselines. In contrast, uniformly reducing dimensionality across all token positions leads to worse performance. These results challenge the common assumption that key and value dimensionality should be uniform across token positions. We further develop an efficient realization of DAR that achieves model quality comparable to the MLA baseline while reducing the average KV-cache size by 68.8\% at 8K context and achieving up to a $1.9\times$ decoding speedup at 128K context on the evaluated models. Our findings suggest a new direction for designing attention architectures that allocate representational capacity according to token distance to reduce KV-cache during inference.

cs.CL↗

From Apparent Search Gains to Selection Failures: Auditing Test-Time Inference in Structured Workflow Planning

Test-time search can improve structured planning, but when methods differ in prompts, post-processing, and selection, it is unclear what an observed gain should be attributed to. We study this question on DisasterBench, where models compose typed multi-tool workflows, by decomposing inference into three observable stages: whether a correct workflow is discovered, whether it survives structural validation, and whether it is finally selected. Comparing independent sampling, a nominal tree-search configuration, and multi-operator search under controlled inference budgets on two model families, we find that (i) the nominal tree-search configuration realizes non-adaptive complete-workflow sampling, and independent sampling with its expansion prompt reproduces its candidate discovery without constructing a tree; (ii) structural validation never removes the last correct workflow in our evaluated candidate pools, whereas majority voting leaves a discovered correct workflow unselected on 10-19% of tasks per pool, with an audit confirming substantive tool-choice or information-flow errors in 57 of 60 examined cases; and (iii) a tool-specification-aware verifier recovers some of these losses but overturns more correct decisions than it fixes. Our results motivate auditing realized rather than nominal inference pipelines, and reporting discovery, validity, and selection separately.

cs.CL↗

BiShield-TEE: On the (In-)Security of Unilateral Weight Obfuscation in On-Device TEE-Shielded LLM Partition

On-device deployment of large language models (LLMs) is increasingly adopted due to latency and privacy requirements, but exposes proprietary models to reverse engineering and intellectual property theft. Trusted Execution Environments (TEEs) provide hardware-enforced isolation for on-device inference, yet most TEE-based parameter obfuscation relies on unilateral matrix transformations that obfuscate only the output side. We identify a vulnerability in this design, termed the Solution Determinacy Property (SDP): the obfuscation secret is uniquely determined, exposing the structure of the true key and enabling attackers to break the protection. Under our attacks, representative methods GroupCover and ArrowCloak are reduced to negligible protection, with attack accuracy reaching over 2.4 times the full-shielding black-box baseline and recovering near-complete model functionality. To address this vulnerability, we propose BiShield-TEE, a bilateral framework that induces Ambiguous Solutions through joint obfuscation of input and output dimensions, hiding the true key among many equally valid candidates. Experiments across diverse models and datasets show that BiShield-TEE confines attack accuracy to the full-shielding black-box baseline, while achieving higher inference efficiency than prior methods on real hardware. Our code is available at https://github.com/laduiw/BiShield-TEE.

cs.CR↗

A Sign Law for Entropy Drift under Sampler Mismatch in RLVR

In reinforcement learning with verifiable rewards (RLVR), whether entropy rises or falls can depend on choices that matter little under exact sampling, such as whether over-length responses are masked or counted as failures. We trace this to sampler mismatch: rollouts are drawn with top-k/top-p truncation or at a temperature the learner does not apply. We prove that a constant advantage shift then changes the entropy of a softmax policy, at first order and in expectation, with a sign fixed by the sampler. Positive shifts sharpen the policy under any sampler whose likelihood ratio to the policy increases with token probability, including top-k, top-p, min-p, and temperatures below one, and flatten it under flatter samplers. In group-based RLVR with token-level losses, the baseline-dependent part of this drift is set by the token-weighted net advantage, which masking and failing can push to opposite signs. The rest comes from a covariance between advantages and the sampler's sharpness gap and feeds back on itself: in a two-level model, a push away from a sharper sampler takes off in finite time, while a pull toward it saturates. On a 2B model, the entropy slope is linear in shifts of the net advantage under truncated sampling and indistinguishable from flat under exact sampling. Token-level importance ratios leave 98% of the drift, and four clipped minibatch updates per batch amplify it sevenfold. In 100-step runs, a net-zero baseline only delays an entropy explosion, and entropy raised at the start keeps rising only under truncation; cancelling the first-order drift at every step keeps entropy bounded without changing the learner. On a 9B reasoning model forked from an uncorrected run, zeroing the net advantage or renormalizing the learner raises held-out accuracy after ten steps by 8.2 and 8.8 points over the uncorrected rule, with no difference under exact sampling.

cs.LG↗

When Safety Fails Collectively: Robust Alignment Against Harmful Fine-Tuning via Joint Weight Perturbations

Harmful fine-tuning can rapidly erode the safety alignment of large language models, creating substantial risks when aligned models are adapted to downstream tasks. Existing alignment-stage defenses use harmful examples and simulated harmful perturbations during safety alignment to prepare models for later harmful fine-tuning, yet the collective effects of harmful weight perturbations remain underexplored. We identify a collective failure pattern: disjoint parts of a harmful perturbation have limited effects in isolation, but their joint application increases the model's harmfulness beyond the sum of their individual effects. Guided by this observation, we propose Joint Weight Perturbations (JWP), an alignment-stage defense that masks different parts of a harmful perturbation estimated from the current model and trains the model to respond safely in the resulting states. This exposes the model to different combinations of harmful perturbations and reduces its sensitivity to their accumulation during subsequent harmful fine-tuning. Experiments across downstream tasks and model families show that JWP offers stronger resistance to harmful fine-tuning than prior defenses while largely preserving task performance.

cs.AI↗

Learning to Program Retrievals: Composing Search Primitives via Reinforcement Learning

Retrieval-augmented language models increasingly use multi-step search to solve knowledge-intensive tasks, yet existing systems typically expose retrieval through fixed pipelines or collections of atomic tools. We introduce Retrieval Program Learning (RPL), a programmable search framework that allows a policy to compose search primitives into executable programs with local state and control flow. This enables adaptive search that can retrieve broadly, process intermediate results, and selectively construct model context without tying retrieval breadth to context consumption. However, the resulting flexibility creates a difficult exploration problem: useful programs may require several coordinated decisions, while reinforcement learning often supplies only sparse feedback. To address this challenge, we propose program mutation, a grammar-constrained and execution-verified procedure that generates local variants of suboptimal programs sampled during RL, converts them into training trajectories, and distills them back into the policy. We evaluate RPL on open-domain question answering and corpus-level reasoning tasks, comparing it with a range of retrieval and reasoning baselines. RPL consistently improves open-domain QA performance across model sizes and in- and out-of-domain benchmarks, with particularly large gains on multi-hop tasks. Notably, a 4B model trained with RPL outperforms all 8B model baselines. On corpus-level reasoning, RPL achieves the highest answer F1 and document-coverage F1 for both model sizes while substantially reducing context consumption and retrieval turns. Our results show that programmable search expands the retrieval strategies that agents can learn, while program mutation provides an effective mechanism for exploring useful program structures beyond direct on-policy sampling.

cs.CL↗

X-Ragent: Tool-Augmented Reinforcement Learning for Evidence-Grounded Chest X-Ray Reasoning

Chest X-ray (CXR) interpretation often requires integrating visual findings with task-specific evidence rather than relying on direct visual recognition alone. While tool-augmented medical agents can supply such evidence via external modules, they often lack explicit mechanisms for learning when auxiliary evidence is needed and for preventing inapplicable quantitative evidence from entering diagnostic reasoning. To address this, we propose X-Ragent, a post-training framework for tool-augmented reinforcement learning that decouples learned auxiliary-evidence seeking from validity-gated quantitative measurement acquisition for evidence-grounded CXR reasoning. We construct Rad-CoT, a 7.7K-sample corpus combining causally ordered tool-augmented trajectories with report-style diagnostic reasoning, and use it to initialize Qwen3-VL-8B via agentic supervised fine-tuning (SFT). We then optimize the policy with Group Relative Policy Optimization (GRPO) using an accuracy-gated reward that applies process-level terms for measurement consistency and auxiliary-tool cost only to correct trajectories, preventing plausible process behavior from compensating for an incorrect diagnosis. Across three CXR benchmarks, X-Ragent achieves the strongest results, including an 8.6-point improvement on the fully held-out ChestAgentBench. By seeking evidence selectively and admitting only valid measurements, X-Ragent supports more faithful and evidence-grounded CXR reasoning.

cs.CV↗

X-AgentRed: Red-Teaming Personal Agents in Evolving Multi-Service Ecosystems

Personal agents increasingly connect to services across the web and applications by using tools. As the effects of their actions accumulate and interact across services, unintended or harmful outcomes can arise when agents fulfill benign user requests. Proactively identifying these compositional failures requires systematically exploring diverse cross-service paths. Existing red-teaming methods often simply prompt LLMs to identify potentially harmful paths from task and environment descriptions or decompose harmful objectives specified by humans or LLMs into seemingly benign action combinations. However, generation is biased toward patterns learned during training, potentially favoring familiar tool combinations and leaving novel cross-service interactions underexplored. To address this, we propose X-AgentRed, a self-evolving red-teaming framework that discovers harmful tool call paths through a graph of tool dependencies and expands and refines them using execution feedback. By analyzing each tool's effects, potential harms, and preconditions for execution, the red-teaming agent constructs a directed graph that captures how one tool's effects can satisfy another's execution preconditions. Recipes sampled from this graph are instantiated as realistic scenarios in which agents can produce harmful outcomes. Execution feedback enables self-improvement by diversifying paths that produce harm through tool substitution and refining preconditions at blocked steps in unsuccessful paths to enable further execution. Across four frontier agentic models, X-AgentRed achieves $1.77\times$ the harmful path rate of STAC and explores a more diverse set of tools, yielding $2.99\times$ the tool coverage on average.

cs.AI↗

On leading Gopakumar--Vafa invariants of local $\mathbb{P}^2$ and leading Betti numbers of moduli spaces of one-dimensional sheaves on $\mathbb{P}^2$

In this paper we use the topological vertex formalism to derive a structural formula for the leading terms of the generating series of Gromov--Witten invariants of local $\mathbb P^2$. Then we give two applications of this formula. First we use this result to derive an explicit formula for the leading terms of the transformed Gopakumar--Vafa invariants of local $\mathbb P^2$. This is an extension of a formula of Guo--Zhou \cite{gz} from degree $2d-5$ to $3d-1$, and shares some similar features with the famous Göttsche--Yau--Zaslow formula. And as the second application, we prove Guo--Wu--Moreira's Conjecture \cite{gw} about the leading Betti numbers of the moduli space of one-dimensional Gieseker semistable sheaves on $\mathbb P^2$.

math.AG↗