arXiv subjects

Yichi Zhang

Publications and source records attributed to Yichi Zhang.

At least 19 recordsLinked to original sources

arXiv2026

Electronic liquid crystals and superconductivity from doping a moiré Mott insulator

Electronic liquid crystals, such as stripe phases and nematics, are self-organized states of matter with spontaneously broken rotational symmetry. They are expected to emerge in a doped Mott insulator as a compromise between short- and long-range Coulomb interactions, and are believed to be intimately connected to the emergence of high temperature superconductivity either as a competing order or an intertwined order. Moiré semiconductors provide a controllable platform to explore the relationship between stripe phases and superconductivity near a Mott insulator. Here we report the generic appearance of stripe phases in a doped moiré Mott insulator and the observation of a stripe to-superconductor quantum phase transition in twisted bilayer WSe2. The stripe phases were revealed by electrical transport studies as a thermodynamic phase transition from a high temperature isotropic phase to a low-temperature anisotropic phase accompanied by a Fermi surface reconstruction. Near the stripe-to-superconductor transition, we also observed an additional thermodynamic phase transition from metallic stripes to pinned insulating stripes. The stripe-to-superconductor transition further exhibits an intriguing self-similarity that highlights the intimate relationship between the two orders. Our results pave the path for studying intertwined orders in moiré materials.

cond-mat.str-el↗

arXiv2026

Every Little Helps: Building Knowledge Graph Foundation Model with Fine-grained Transferable Multi-modal Tokens

Multi-modal knowledge graph reasoning (MMKGR) aims to predict the missing links by exploiting both graph structure information and multi-modal entity contents. Most existing works are designed for a transductive setting, which learns dataset-specific embeddings and struggles to generalize to new KGs. Recent knowledge graph foundation models (KGFMs) improve cross-KG transfer, but they mainly exploit structural patterns and ignore rich multi-modal signals. We address these gaps by proposing a token-based foundation model (TOFU) for MMKGR, which exhibits strong generalization across different MMKGs. TOFU discretizes structural, visual, and textual information into modality-specific tokens. TOFU then employs a hierarchical fusion architecture with mixture-of-message mechanisms, aiming to process these tokens and obtain transferable features for MMKGR. Experimental results on 17 transductive, inductive, and fully-inductive MMKGs show that TOFU generally outperforms strong KGFM and MMKGR baselines, delivering strong performance on unseen MMKGs.

cs.CL↗

arXiv2026

Correlation-sensitive magic-state distillation from spin-squeezed neutral atoms

We show that interblock correlations can reverse magic-state distillation even when every input exceeds the threshold for independent states. For the five-to-one Bravyi-Kitaev protocol with independent local Clifford twirls, we express the output in terms of measurable target-axis moments. Fourth-order correlations determine acceptance, while third- and fifth-order correlations determine the output polarization. This relation yields a direct sign test for distillation gain and opposite gains for inputs with identical single-qubit and two-qubit reduced states. We connect this mechanism to neutral-atom resource preparation through finite-range Ising dynamics. Block separation suppresses detrimental interblock coupling, and locally addressed Clifford phases release interaction-generated magic from a spin-squeezed precursor by changing the symmetry selected by code projection. A protocol that retains accepted source outputs achieves positive gain above a source-contrast threshold set by circuit errors and storage dephasing. The same moments specify a finite-sample experimental diagnostic. These results connect the correlation structure of a resource state to its preparation, acceptance, and distillation gain.

quant-ph↗

arXiv2026

Xetrieval: Explaining Dense Retrieval through Feature Elicitation

Dense retrieval enables effective semantic matching, yet the semantic factors underlying its relevance scores remain opaque. Sparse autoencoders (SAEs) support feature-level explanations by decomposing dense embeddings into sparse semantic features. However, the document semantics underlying a retriever's relevance scores may not be sufficiently explicit in the embeddings to be recovered as interpretable SAE features. We introduce Xetrieval, a framework for explaining dense retrieval through feature elicitation. Feature Elicitors make retrieval-relevant aspects of document meaning more explicit through auxiliary views, facilitating their identification as interpretable SAE features. A shared sparse dictionary learns from these varied representations to express query--document relations through describable features. We assess their influence on relevance scores through interventions on the original document embeddings. Experiments show that Xetrieval yields more interpretable features than an SAE baseline and identifies feature sets with a stronger influence on the retriever's scores. Task-level feature modulation further demonstrates more effective retrieval steering. These findings highlight the value of feature elicitation for relevance attribution and retrieval steering.

cs.AI↗

arXiv2026

Extended strange metal and superconductivity in partially filled moiré Hubbard bands

Moiré semiconductors provide a physical realization of the Hubbard model. Bandwidth-tuned Mott transitions and complex phase diagrams mimicking those of high-temperature superconductors have been demonstrated near a weak Mott insulating state at half-band-filling. The platform also provides an opportunity to explore the physics of a partially filled Hubbard band at doping densities far away from the Mott insulator, a regime that has remained largely unexplored. Here we report the emergence of an extended strange metal (ESM), in both doping density and twist angle, in a partially filled Hubbard band of twisted bilayer WSe2. The ESM is in the clean limit, persists down to about 50mK, and spans over multiple symmetry-breaking ground states; it also appears only at small twist angles, where electronic correlations are strong. These observations distinguish the state from the reported strange metals associated with quantum criticality and favor the viewpoint that it is an intrinsic state of matter. We also report the observation of a new superconducting dome being developed from the ESM but far away from both the Mott insulating state and the van Hove singularity of the material. Our work opens the door to exploring the new regime of partially filled Hubbard bands in the clean limit.

cond-mat.mes-hall↗

arXiv2026

Rethinking Quality in Large Language Model Routing: Security-Conditioned Model-and-Budget Selection

Large language model (LLM) routers reduce serving cost by choosing which model to call and how many output tokens to grant it for each query. Existing routers rank model-budget actions using a single quality metric. This does not work for mixed workloads. For ordinary queries, a good answer is correct and useful (utility); for security-sensitive queries, a good answer refuses harmful requests or resists malicious instructions (security). These two objectives prefer different models and budgets. We propose SCRouter, a router that learns when to prioritize security and how to allocate the token budget. For each model-budget pair, it estimates the probability that the response finishes within budget. It uses this probability to predict utility and security separately. A small probe determines how to weight these two predictions for each query. SCRouter then selects the model and budget by balancing the weighted utility and security against cost. We also build SCRouterBench, a benchmark of 38,869 utility and security queries spanning nine security dimensions, each answered under all 144 combinations of 12 models and 12 budgets, plus 2,753 out-of-domain queries. SCRouter improves the baselines by 13-29 points in-domain and 19-23 points out-of-domain.

cs.AI↗

arXiv2026

Share the Symbol, Not the Vector: Training-Free Fact Communication Across Language Models

Multi-agent systems allow language models to solve tasks collaboratively by sharing complementary information. These systems typically communicate through text, while recent approaches also explore exchanging internal model states. However, sharing these states across architectures often requires learned translators or assumptions about model compatibility. We introduce SYNC (Symbol-Indexed Neural Communication), a word-indexed interface that connects heterogeneous models without fine-tuning or a learned translator. From descriptions only it can see, the sender selects words from a shared list and sends their indices for each object, without seeing the questions. The receiver writes the described facts into its hidden states using its own word directions, vectors fitted once from its activations. We test selective use by swapping one attribute between two objects and requiring correct answers about both changed and unchanged attributes before and after the swap, four answers in total. We choose an early write layer on four development receivers and then fix it. With three different senders and nine newer receivers from five families, four-answer success on new scenes reaches 72.0-98.3%, up to 96 percentage points above middle-layer writing, and the two largest receivers match plain text. A task-free probe predicted in advance which depths each new receiver can use (45 of 45 cases). Without retuning, messages listing every fact keep 89.6-100% success under new layouts, and sender messages reach 23.9-64.0% on a landmark task with multi-token symbols, below text but from 0% without a message. These results show that shared word indices and receiver-local directions can support selective fact use across model families without additional training.

cs.CL↗

arXiv2026

TacClaw-Infinity: A Large-Scale Vision-Tactile-Language-Action Dataset for Dexterous Manipulation

Large-scale robot learning increasingly relies on diverse human demonstrations, yet tactile manipulation data remain difficult to scale, and existing multimodal datasets often lack consistent quality control and structured supervision. We present TacClaw-Infinity, a large-scale structured Vision-Tactile-Language-Action (VTLA) dataset for contact-rich manipulation, collected with TacClaw, a tactile-enabled UMI-style handheld interface that records synchronized visual observations, fingertip tactile signals, and end-effector trajectories without requiring a target robot during data collection. TacClaw-Infinity is built through a distributed worldwide collection across homes, laboratories, industrial workplaces, retail stores, and outdoor environments including tasks contributed by academic and industrial collaborators. Together, these efforts yield over 10,000 hours, 1 million episodes, and 1 billion frames spanning more than 1,000 task types. We develop a scalable pipeline for trajectory reconstruction, multimodal quality control, embodiment-aware filtering, and structured annotation, covering contact-centric segmentation, subtask description, and spatial grounding. We validate TacClaw-Infinity at both the data and model levels: At the data level, we assess the collection interface, the action-space compatibility of the curated demonstrations, and the structured annotations, while for the model level, we demonstrate cross-embodiment transfer across three heterogeneous robots without robot-specific demonstrations.

cs.RO↗

arXiv2026

Secure Speculative Decoding for Large Language Models

Speculative decoding accelerates inference for a large language model (LLM), referred to as the \emph{target model}, by first using a smaller model, referred to as the \emph{draft model}, to generate candidate tokens and then verifying them with the target model for acceptance or rejection. Prior studies primarily focused on the efficiency-utility trade-off of speculative decoding, e.g., lossy speculative decoding, leaving its security implications largely unexplored. In this work, we bridge this gap by providing the \emph{first} systematic study of the security implications of speculative decoding. Through a large-scale measurement study, we reveal a pronounced security-utility asymmetry: across a wide range of lossy speculative decoding methods, improvements in inference efficiency come at a disproportionately high cost to security, with attack success rates for jailbreak and prompt injection attacks increasing much faster than utility degrades. We then propose SecureSD, a new theory-guided speculative decoding method that enhances security while maintaining efficiency and utility. Specifically, our theoretical analysis reveals that security degradation primarily originates from the early tokens generated by the draft model. Motivated by this insight, SecureSD applies a stricter verification criterion to draft-model tokens at early decoding positions. Extensive experiments on both security and utility benchmarks demonstrate that SecureSD significantly improves security while preserving efficiency and utility compared to existing speculative decoding methods.

cs.CR↗

arXiv2026

ErrOPD: When Does On-Policy Distillation Transfer Teacher Errors?

A distilled student can copy its teacher's mistakes along with its skills. On-policy distillation (OPD), used in recent post-training recipes, has the teacher grade every token of the student's own outputs. Which teacher errors does it pass on? We build ErrOPD, a testbed whose teachers are fine-tuned to close a JSON list one item early, an error their students almost never make. The answer depends on the KL direction: across four model families, forward-KL OPD transmits the error within the training budget (+13-23 points), while reverse-KL OPD at our reference learning rate adds under 0.05 points. A duplicated ID and, where the decision arises, two SQL errors show the same split. The reverse-KL update scales with the student's own probability of a token, and an intervention tests this: giving reverse KL forward KL's gradient magnitude at the error decision alone makes it transmit on two model pairs. Reverse-KL OPD can delay such errors rather than block them: some runs adopt the error with longer training, and Qwen3 students that already make it, even rarely, adopt it within the budget. Because greedy evaluation on Qwen3 misses the inherited error, distilled students should also be audited on sampled outputs.

cs.LG↗

arXiv2026

Lost in Context Compaction: When LLM Systems Drop Side Constraints

When the context window is under pressure, LLM systems compact prior context to continue ongoing tasks. We study Side Constraints (SCs), user-issued instructions such as "do not delete any emails until I confirm," that constrain an LLM's behavior across subsequent turns but can be silently dropped during compaction. In an inbox experiment, an agent deleted emails without approval in 30/100 sessions after compaction, compared with none before compaction. To systematically evaluate SC preservation, we introduce CONPACT, an evaluation suite measuring semantic retention and downstream compliance across three scenarios: multi-turn chat, multi-turn tool use, and long-horizon research. The evaluated compactors retain only 17% of injected SCs on average. Loss occurs across compactors and settings, with retention varying by compactor, context length, SC framing, and injection location. We propose an SC extractor that maintains constraints separately from the task summary and appends them to the compacted context, achieving over 90% retention across all three scenarios without modifying the compactor or LLM. Code and data are available at https://github.com/ZhiqiEliWang/compaction-integrity.

cs.CL↗

arXiv2026

Augmented Roothaan-Hall Hessian Applied to Spin-Restricted Open-Shell Density-Functional Theory

We generalize the augmented Roothaan-Hall (ARH) Hessian formalism to the self-consistent field (SCF) optimization of spin-restricted open-shell (RO) wavefunctions, encompassing high-spin, low-spin, and two-determinant electronic states. A detailed ARH formulation is presented. We demonstrate that ARH is a highly efficient optimization algorithm for rapidly identifying accurate SCF solutions, primarily owing to its systematic construction of an effective Hessian, particularly in the case of Euclidean quadratic energy functions. The ARH is built upon a universal energy formulation, including grid-based integration, for spin-restricted closed-shell, spin-unrestricted and RO density functional theory (DFT), thereby unifying and simplifying their numerical implementation. The performance of the present method is evaluated using two benchmarking studies. First, for a series of iron-sulfur clusters exhibiting different spin states, which represent notoriously challenging SCF problems, the ARH algorithm demonstrates superior convergence efficiency relative to L-BFGS and truncated Newton methods, requiring much fewer RO-SCF iterations to achieve convergence. Second, the ARH approach tends to avoid convergence to higher-energy stationary points in two-determinant RO-SCF calculations for singlet excited states of selected photoactive compounds. Finally, an application of the ARH-based RO-SCF is illustrated by an investigation of the mechanistic origin of the spin-crossover phenomenon in Ni(II)-porphyrin complex utilized as a contrast agent.

physics.chem-ph↗

arXiv2026

Unifying Distributional Training for One-Step Visual Generation

Distributional training provides collective supervision for one-step visual generation by matching real and generated features in frozen representation spaces. We introduce a unified theoretical framework that separates distribution modeling from matching discrepancy and connects global objectives to pointwise feature updates through Wasserstein gradient flow. Under this framework, FD-Loss and Gaussian-kernel Drifting are recovered through Gaussian optimal transport and kernel-density-based KL matching, respectively. The framework motivates MGFlow, which models feature distributions with Gaussian mixtures at an adjustable granularity between global moments and sample-based representations. MGFlow supports both optimal transport and score-based matching, and couples mass-constrained sample assignment with paired component updates to address mode collapse that mixture expressivity alone does not resolve. On ImageNet $256\times256$, MGFlow substantially surpasses the FD-Loss baseline, achieving state-of-the-art results with 1.45 $\mathrm{FDr}^6$ on pMF-H and 1.64 on JiT-H. For text-to-image generation, MGFlow post-trains FLUX.2 [klein] 4B into a one-step generator that outperforms the original four-step model on both GenEval and PickScore.

cs.LG↗

arXiv2026

EgoTools: Towards Tool-Centric Reasoning in Real-World Egocentric Videos

Real-world embodied tasks, from everyday activities to professional procedures, require agents to act under physical constraints while tracking evolving object and task states. Tool use sits at the heart of such tasks, as many everyday and professional activities are tool-mediated. Understanding them requires reasoning about affordances, hand-tool-object geometry, procedural progress, and causal effects on target objects. Yet despite strong performance on perception-oriented video tasks such as captioning and general video QA, current multimodal video models remain limited in this form of tool-centric embodied reasoning. Progress in this direction has been limited by the lack of real-world egocentric data and diagnostic benchmarks. To address this gap, we introduce EgoTools, the first comprehensive suite for egocentric tool-use understanding. It consists of two complementary components: EgoTools-Data, a large-scale corpus of 100 hours of tool-centric egocentric recordings with synchronized audio, dense captions, reasoning-heavy narrations, and supplementary 3D information; and EgoTools-Bench, a diagnostic benchmark of 1,000 QA pairs across four tracks that cover tool-use understanding from perception and geometry to procedure and causal reasoning. Experimental results show that current models still struggle to ground tool use in visual evidence: Gemini-3.1-Pro achieves 66.9% overall accuracy but only 51.7% on Perception & Grounding. Beyond evaluation, we validate EgoTools-Data as a training resource. On the full 1,000-question benchmark, full supervised fine-tuning improves Qwen3-VL-8B-Instruct from 50.0% to 60.9%, under strict source-video separation. Together, these results establish EgoTools as a unified resource for both training and diagnostic evaluation of real-world egocentric tool-use understanding.

cs.CV↗

arXiv2026

hacktrace: behavior-supervised detection of reward hacking during code generation

A coding agent can earn a passing grade by fixing its code, or by deleting the test that exposes the bug. Detecting such reward hacking requires recognizing attempted shortcuts, including those that fail. We release 173,561 annotated multi-turn coding trajectories from Qwen3-8B and show that supervising shortcut behavior independently of exploit success substantially improves detection. We introduce HACKTRACE, a behavior-supervised monitor that reads the internal states the agent already computes while generating code. Reusing these states enables monitoring before a turn is complete, without additional language-model tokens or passes. Combining this evidence with static features of the final files achieves a mean per-problem AUC of 0.997 with 8 ms of monitoring overhead, improving both accuracy and latency over monitors that run the model again on an honesty question and answer. The same generation states also provide an inexpensive monitoring signal for reinforcement learning. With strong GRPO penalties, HACKTRACE reduces the cheating share of passing solutions from 82-91% to 1-5%, while retaining honest, correct solutions and maintaining high detection accuracy as the policy evolves. Our results show that both the supervision target and the source of monitoring evidence matter for turning accurate detection into a useful training signal.

cs.AI↗

arXiv2026

AneumoBench: A Source-Linked Benchmark for Synthetic-Geometry Transfer in Aneurysm CFD

Scientific machine learning uses simulation data to train surrogate models for fast physical-field prediction across geometries. Local shape editing can expand limited geometry collections, but whether its variants improve prediction on unseen geometries, and how to allocate them across sources, require controlled evaluation. We introduce AneumoBench, a dataset and benchmark linking 401 source aneurysm geometries to 9,693 locally edited descendant records, with computational fluid dynamics (CFD) fields computed on both. It contains 80,752 steady velocity-pressure cases across eight inlet conditions and 9,715 transient sequences of velocity, pressure, and wall shear stress (WSS). Each sequence contains 100 frames sampled at 0.01-s intervals from a 1-s cardiac cycle. Mesh, point, and voxel interfaces support steady field prediction and WSS forecasting from four observed frames. With family-disjoint splits, we compare source-only training, descendant training, and descendant pretraining followed by source fine-tuning across nine architectures on 79 held-out sources. Under the reported schedules, two-stage training lowers steady-field and reset-window WSS errors relative to source-only training. With the number of sampled fields and training updates fixed within each comparison, GraphSAGE benefits from descendant training and from distributing a fixed number of descendants across more sources. For WSS, reset-window gains do not consistently persist through 96-step rollout, and lower trajectory error need not improve cycle-level shear metrics or hotspot localization. These data and protocols enable researchers to compare descendant selection and training strategies on the same unseen source geometries.

eess.IV↗

arXiv2026

Quantifying and Correcting Measurement Error in LLM-Generated Classifications: A Beta-Binomial Latent Class Model for Correlated Repeated Judgments

Large language models are increasingly used to classify text at scale. Because their outputs are stochastic, it is common to query a model several times per item and aggregate the results, and latent class models offer a principled way to convert such repeated ratings into estimates of outcome prevalence and of the model's error rates. These models assume that repeated ratings of the same item are conditionally independent given the true label. We examine this assumption using 52,500 judgments from seven open-weight models ranging from 1B to 32B parameters, collected under three prompt conditions with ten samples per item on a public benchmark with human labels. On the balanced split of the benchmark, intraclass correlations lie between 0.50 and 0.93, so ten ratings carry the information of between 1.07 and 1.81 independent ratings. Under this correlation, a latent class model with a binomial likelihood is both miscalibrated and biased. Its nominal 95% intervals for the false acceptance rate contained the value computed from human labels in none of 42 settings, and it underestimated the total error rate in all 42. We propose a beta-binomial latent class model with a single correlation parameter, establish its identifiability and that of extensions to covariates, treatment effects and partially labeled data, and show that on the balanced split it raises the coverage of nominal 95% intervals for the prevalence and the false acceptance rate to 0.90. Recovery of the latent state is governed by the gap between the model's acceptance rates on truly positive and truly negative items, which equals Youden's J and predicts recovery almost monotonically (Spearman correlation 0.96). When this gap is small, a labeled subset improves estimates of prevalence but not the classification of individual items. Simulation studies designed around the identification conditions support these findings.

stat.ME↗

arXiv2026

SIMLIFE: Pattern Understanding for Long-Horizon Human-Agent Partnership

Understanding humans over long horizons requires agents to infer not only what people need in the moment, but also how routines form, why they repeat, and when they change. We introduce SimLife, a scalable platform for simulating long-term household life with rich visual observations, ground-truth action logs, and synthetic dialogues with audio. Built on SimLife, SimLife-BP evaluates long-context pattern understanding: the ability to infer latent behavioral rules from weeks or months of everyday observations. The benchmark contains 106 episodes averaging 15.49 hours and 38.57 in-game days, and 1,439 question-answer pairs. Each task probes direct, counterfactual, noisy, and inverse reasoning under different levels of rule hints. Evaluating frontier models and architectures, we find that current models often achieve surface-level prediction without comprehensive rule understanding, rely on frequency-based heuristics rather than if-then reasoning over evidence, and struggle to adapt when behavioral patterns change. These findings suggest that long-context pattern understanding remains a major bottleneck for future embodied agents, while SimLife opens a broader space for studying memory, personalization, adaptation, and long-horizon planning in everyday human-AI interaction.

cs.AI↗