Searcharxiv⌕ Search

arXiv subjects

Yifan Zhang

Publications and source records attributed to Yifan Zhang.

At least 91 records · Page 5Linked to original sources
arXiv2026

MathGen: Revealing the Illusion of Mathematical Competence through Text-to-Image Generation

Modern generative models have demonstrated the ability to solve challenging mathematical problems. In many real-world settings, however, mathematical solutions must be expressed visually through diagrams, plots, geometric constructions, and structured symbolic layouts, where correctness depends on precise visual composition. This naturally raises the question of whether generative models can still do so when the answer must be rendered visually rather than written in text? To study this problem, we introduce MathGen, a rigorous benchmark of 420 problems spanning seven core domains, including 350 Clean-Scene problems and 70 paired Open-Scene problems. Each problem is evaluated under a Script-as-a-Judge protocol with problem-specific verification criteria implemented through reusable executable scripts for deterministic and reproducible evaluation. Experiments on representative open-source and proprietary text-to-image models show that mathematical fidelity remains a major bottleneck: even the best closed-source model reaches only 53.7% overall accuracy, while open-source models achieve just 1--11%, often near 0% on structured tasks, particularly those requiring precise geometric and functional rendering. Overall, current T2I models remain far from reliable at even elementary mathematical visual generation.

cs.CV↗
arXiv2026

Beyond Field Accuracy: Two-Axis Diagnosis of Inverse-PINN Parameter Error

Inverse physics-informed neural networks (PINNs) can reconstruct a field accurately while returning an incorrect physical parameter. We introduce a two-axis post-training diagnosis that separates finite-sample resolution under a specified observation-and-estimation protocol from the signed parameter preference encoded by the final learned field and residual metric. The first axis repeatedly fits noisy observations with a matched forward estimator. At known synthetic truth, the second freezes the field and residual view and computes a local score displacement toward a nearby residual-profile minimum. Endpoint consistency then tests whether joint training delivers that preference under the same final view. Across three synthetic one-dimensional, scalar-parameter PDEs, matched-forward mean absolute relative error ranges from 2.34 percent to 17.46 percent. The displacement tracks frozen-profile minima across locked seeds, architectures, and fresh-noise retraining (r from .945 to .982), and it tracks delivered signed log-error in 240 fresh-noise RBA runs (r = .994; 237/240 correct directions). A coupled two-parameter Darcy check validates the full matrix calculation. The axes are complementary diagnostic coordinates, not additive error components or a deployable oracle-free estimator. Together, they route follow-up work toward observations, residual evidence, or endpoint delivery.

cs.LG↗
arXiv2026

Handling covariate shift by model averaging

Distributional mismatch between the data used to construct a statistical procedure and the population to which it is ultimately applied is pervasive in modern data analysis. We study covariate shift, a fundamental instance of this problem, and develop an adaptive importance-weighted model averaging method for prediction when labeled observations are available from a source distribution, whereas only unlabeled covariates are observed from the target distribution. Procedures fitted directly to the source sample generally optimize prediction risk under the source distribution and may therefore be suboptimal for target prediction. Importance weighting by the density ratio between the target and source covariate marginals provides a natural correction, but a small number of large density-ratio values can substantially inflate the variance of the resulting estimator in finite samples. We address this bias-variance trade-off by treating the degree of importance-weighting correction as a source of model uncertainty. Specifically, we construct a family of adaptive importance-weighted least-squares estimators by raising the estimated density ratio to a range of exponents, with the endpoints corresponding to ordinary least squares and standard importance-weighted least squares, and form a data-driven average over these candidates. Under model misspecification, the proposed model averaging estimator is shown to be asymptotically optimal relative to the infeasible best convex combination of the candidate estimators. Under correct specification, a diverging penalty is shown to make the selected weights concentrate near the ordinary least-squares endpoint. Simulations and a real-data application show that the proposed method achieves competitive target-prediction performance across the settings considered.

stat.ME↗
arXiv2026

AmbSentry: Mitigating Sensing Eavesdropping in ISAC Systems by Harnessing Ambient IoT Devices

Integrated sensing and communication (ISAC) has emerged as a pivotal paradigm for 6G networks, enabling the synergistic convergence of spectral and hardware resources to maximize system efficiency. However, the inherent openness of wireless transmission exposes ISAC systems to critical security risks, particularly regarding the privacy of the sensing information. Unauthorized sensing eavesdroppers can extract sensitive target parameters (e.g., range and velocity) by directly estimating open sensing echo channels, rendering traditional data-based protection techniques ineffective. To mitigate this threat, this paper proposes AmbSentry, an ISAC system that prevents the leakage of sensing information to sensing eavesdroppers by harnessing naturally distributed passive ambient IoT (AIoT) devices. Specifically, these AIoT devices are strategically configured to act as cooperative jammers and ghost targets, introducing controllable interference into the sensing environment. Based on the proposed system, we formulate a joint optimization problem to maximize the integrated sidelobe level at the eavesdropper under quality-of-service (QoS) constraints, thereby degrading sensing eavesdropping performance while maintaining sensing and communication performance for legitimate receivers. Since the problem is non-convex, we further develop an efficient iterative algorithm to cooperatively design the transmit beamforming at the base station and the reflection modulations of the AIoT devices based on Dinkelbach transformation and block coordinate descent methods. The detailed results also demonstrate that AmbSentry significantly enhances sensing security, allowing the legitimate sensing receiver to achieve a 14-dB SNR advantage in detection probability and a hundred times lower estimation error compared to the eavesdropper.

cs.CR↗
arXiv2026

LoKiFormer: Locality-aware Attention with Decoupled Knowledge Memory for Efficient Large Language Model Pretraining

Large language models (LLMs) have achieved remarkable breakthroughs across various applications. However, their architectures remain inefficient in pretraining due to two main limitations: (i) self-attention lacks an explicit inductive bias for locality, leading to redundant modeling of sequence-internal local information; (ii) mixture-of-experts (MoE) implicitly couples knowledge storage with computational pathways, hindering flexible access to sequence-external global knowledge. To overcome these limitations, we propose LoKiFormer, a novel LLM architecture that augments the standard decoder with two dedicated modules: 1) Local Fusion Attention (LFA), which incorporates a convolutional fusion to attention, explicitly capturing local patterns and allowing the attention to operate on more informative representations; 2) Knowledge Memory Module (KMM), which introduces a parametric key-value memory that explicitly stores global knowledge in addressable slots, decoupling storage from computation and enabling direct knowledge retrieval. Together, these modules enable LoKiFormer to achieve more efficient and effective integration of information at both levels. Experimental results show that LoKiFormer converges 1.33x faster in pre-training than baseline models, underscoring its superiority over existing LLM architectures.

cs.LG↗
arXiv2026

Multi-Agent Closed-Loop Reasoning for Organic Structure Elucidation from Multimodal Spectra

Following the molecular discovery and synthesis revolutions, scalable automated structure elucidation from routine spectroscopic data remains an outstanding challenge. Despite decades of computational efforts, no existing system achieved reliable reasoning over unseen spectra. Here, we propose MACROS, a multi-agent system automating structure elucidation by emulating expert iterative hypothesis-testing. Trained on 100M simulated and 1.6M experimental spectra-molecule pairs, it natively supports arbitrary combinations of routine spectroscopic techniques. It achieves unprecedented zero-shot generalization to diverse real-world samples, correctly identifying synthetic compounds, natural products and metabolites above 500 Da with 1D NMR. Remarkably, MACROS spontaneously recovers textbook spectroscopic correlations from unassigned data and exhibits emergent chemical intuition such as a ring-first parsing preference, learning fundamental chemical principles rather than memorizing database patterns. MACROS augments chemists via collaboration to deliver sixfold faster, 40% more accurate elucidation. MACROS establishes a scalable foundation for fully automated structure elucidation, and catalyzes accelerated molecular discovery toward autonomous laboratories.

physics.chem-ph↗
arXiv2026

StreamFlow: Dynamic Memory Flows for Streaming Video Understanding

Streaming video understanding requires multimodal large language models (MLLMs) to preserve relevant evidence from continuously evolving streams under strict causality and bounded memory. Yet existing paradigms remain limited: model-based methods require intrusive backbone updates, while memory-based methods expend substantial visual-encoding computation on temporally redundant content and rely on rigid access to visual history. To address these limitations, we introduce StreamFlow, an efficient visual memory framework that enables dynamic, on-demand access to historical visual information. StreamFlow combines a lightweight, dynamics-aware mid-term memory that filters temporal redundancy before visual encoding with a latent long-term memory that consolidates historical video content into visual latents accessible to subsequent reasoning. During generation, an attention-guided retrieval mechanism injects relevant visual latents when the model's reliance on visual evidence weakens. StreamFlow achieves state-of-the-art streaming video understanding performance, reaching 67.73% overall accuracy on StreamingBench, while also delivering strong performance on offline long-video benchmarks. Relative to the vanilla setting, it improves the visual attention score (VAS) by 59.1% while reducing end-to-end latency and peak memory by 50.4% and 21.1%, respectively, enabling more visually grounded and efficient reasoning.

cs.CV↗
arXiv2026

Coupling Planning with Episodic Memory in LLM Agents for Software Issue Resolution

Resolving a real software issue with a large language model (LLM) agent is a long repair episode, often tens to hundreds of steps spanning exploration, hypothesis, implementation, and verification. Success depends on both the base model's local reasoning and the agent's ability to maintain an evolving plan and remember observations across phases. Existing repository-level agents typically strengthen planning or memory in isolation, leaving long trajectories vulnerable to stale evidence, repeated failed edits, and verification inferred from the agent's own claims instead of execution evidence. We present PMCoder, an issue-resolution agent that couples a hierarchical phase planner with episodic memory. The coupling is bidirectional: the current plan phase conditions memory retrieval, while memory-derived trajectory statistics inform stuck detection and replanning. When available, issue-reproduction verdicts ground verification progress in execution evidence rather than self-reported completion. On SWE-bench Verified, PMCoder resolves an average of $25$ more cases ($+5.0$pp) than a harness-matched baseline, with gains persisting even where the reproduction gate never fires. Further Verified-500 evaluations show the same positive direction across Claude Haiku 4.5, DeepSeek-V4-Flash, and an OpenHands port, with at least $14$ additional resolved cases ($+2.8$pp). Separately, evaluation on TerminalWorld's official sample suggests that the plan-memory substrate transfers beyond issue reports. Ablation and trajectory analyses show where the gains come from: coupling planning and memory outperforms either component alone and reduces repeated failed actions, empty-patch exits, and context-window exhaustion.

cs.SE↗
arXiv2026

DeepLoop: Depth Scaling for Looped Transformers

Looped Transformers scale sequential computation by applying a compact stack of physical blocks for multiple rounds, increasing unrolled depth without increasing stored parameters. This reuse changes the residual-scaling problem: in an untied Transformer, each residual branch receives and applies its own parameter update, whereas in a looped Transformer one shared update aggregates gradients from repeated visits and is read back by those same visits in the next linearized forward pass. We formalize this tied-depth effect through a first-order perturbation bound controlled by a visit-alignment coefficient $κ_R$. The bound recovers the DeepNorm exponent when visits decorrelate, but in the conservative aligned regime it requires the exponent to increase from $1/4$ to $1/2$ as loop count grows at fixed physical depth. The resulting method, \textbf{DeepLoop}, keeps the Post-LN DeepNorm architecture and sets $α=(2N)^{1/2}$ and $β=(8N)^{-1/2}$ for unrolled depth $N$. On GPT-style looped language models at GPT-2 small and GPT-2 medium scale, DeepLoop is neutral when no physical block is revisited and improves validation loss and downstream accuracy once recurrent depth is activated. These results show that stable recurrent depth requires residual scaling rules that account for parameter visits, not only nominal layer count.

cs.LG↗
arXiv2026

Meta Prompting for AI Systems

We introduce Meta Prompting (MP), a framework that emphasizes the formal structure of a task rather than content-specific worked examples. We give a categorical formalization in which a functor maps typed task transformations to typed prompt transformations. Functoriality encodes preservation of identities and composition, but it does not by itself guarantee semantic correctness. We extend MP to Recursive Meta Prompting (RMP), in which a proposer model generates candidate prompt edits, a validator enforces the edit schema, and an executor model uses the resulting prompt. We model the accumulation of valid edit scripts with the Writer monad and their application by a monoid action on prompts. This construction makes accumulated edits independent of parenthesization; it does not guarantee convergence or improved task accuracy. Empirically, a Qwen-72B base model guided by example-free meta-prompts attains 46.3% pass@1 on MATH and 83.5% accuracy on GSM8K. Separately, an MP-CR agent synthesizes a tool-executed Game of 24 solver; the original experiment reports 100% success on 1,362 instances, subject to exact re-verification from the corresponding input and per-instance outputs. The batch amortizes one LLM response across all puzzles, so its token accounting is not protocol-matched to per-instance few-shot or search-based methods.

cs.AI↗
arXiv2026

MUSE: A Heterogeneity-Aware Multimedia Search Engine for Mobile SoCs

On-device multimedia retrieval is vital for smartphones, enabling applications like cross-modal semantic search and multimodal personal AI agents. However, realizing efficient retrieval on mobile SoCs remains a critical systems challenge. Unlike text-based search, multimedia applications depend on high-dimensional cross-modal embeddings (often exceeding 1024 dimensions) and continuously expanding media streams (e.g., video lifelogs and ambient audio). Mobile devices must process these intensive workloads under strict latency, energy, and memory constraints. Existing vector retrieval systems, designed primarily for static server datasets, are poorly suited for the dynamic execution models of mobile SoCs, where interactive queries must coexist with continuous background media ingestion and index maintenance. To address these challenges, we propose MUSE, a MUltimedia Search Engine comprehensively co-designed for mobile SoCs. MUSE tackles these issues through an architecture-index co-design. It introduces a three-stage asynchronous execution pipeline that overlaps DMA transfers, NPU matrix computation, and hardware-vector data adaptation to overcome the memory wall of high-dimensional embeddings. Furthermore, it redesigns the Inverted File Index (IVF) and introduces workload-aware heterogeneous scheduling across the CPU, GPU, and NPU for hybrid interactive-and-ingestion workloads. Evaluated on Snapdragon 8-series SoCs using real-world multimodal datasets, MUSE improves query throughput by up to 1.4x at matched recall, achieves up to 7x faster index construction, and delivers up to 6x higher insertion throughput under concurrent streaming. Crucially, its accelerator-native design ensures strict physical compliance, capping peak device temperatures at 38°C and reducing total energy consumption by up to 4.6x compared to CPU-bound baselines.

cs.DC↗
arXiv2026

MOT-SR: Multi-Objective Tool-Augmented Scientific Equation Discovery with Large Language Models

Symbolic Regression (SR) aims to discover analytical equations from observational data and plays a central role in scientific modeling. While recent Large Language Model (LLM) based approaches show promise, they face two limitations. First, they lack data analysis mechanisms for uncovering variable dependencies, which reduces the efficiency of equation discovery. Second, most methods rely on single-objective evaluation focused solely on fitting error. This neglect of structural complexity and generalization often causes models to converge prematurely to local optima, limiting their ability to explore the broader equation space. We propose Multi-Objective Tool-augmented Symbolic Regression (MOT-SR), a unified framework that integrates external analytical tools to extract structural priors and guide equation generation, while jointly optimizing for accuracy, complexity, and generalization via a multi-objective evaluation module that maintains a dynamic Pareto front. MOT-SR employs two collaborative LLM modules: a Meta Strategy Generator, which selects tools and synthesizes structural optimization strategies based on Pareto-optimal equations, and an Equation Generator, which produces new candidate equations accordingly. The system operates in a closed-loop manner, continuously refining both strategies and equation structures. Across 40 standard tasks, MOT-SR outperforms existing SR methods in accuracy, generalization, and efficiency. We further validate MOT-SR on extreme mass-ratio inspiral (EMRI) orbital modeling, an important problem in space-based gravitational-wave astronomy where small local errors can accumulate substantially over long-term evolution. The discovered interpretable correction achieves the lowest trajectory-level integration error on held-out configurations. These results demonstrate the potential of MOT-SR to enable reliable modeling of long-horizon scientific dynamics.

cs.LG↗
arXiv2026

DarwinX: Evolving Agent Harnesses Through Natural Selection

An LLM agent's capability depends not only on model weights but on its harness: prompts, tools, skills, and control flow. Self-improvement loops already edit harnesses, yet single-lineage search is path-dependent and local wins often regress other tasks. We introduce DarwinX, which treats self-evolution as selection over a population of harnesses with the model frozen: a preserve-and-extend contract admits only variants that extend coverage without regressing, an archive keeps alternative lineages for recombination, and failure-, teacher-, and self-derived evidence share one edit interface. Fitness comes from each benchmark's own verifier: no gold solutions, no hand-picked winners. Across four benchmarks that progressively separate the evolution signal from the test, one loop adds about 17 points on average: Terminal-Bench 2.1 rises +7.7 to 83.2% on a matched base and to the verified frontier at 84.7% on a stronger one; TerminalWorld's held-out split reaches 68.3%, ahead of every off-the-shelf agent; WebArena-Infinity real-task pass@1 rises from 43.5% to 93.0% audit-clean; and a Terminal-Bench 2.1 harness transfers unchanged to SWE-bench Verified. What evolves is general agent competence, not benchmark-specific patches, so it survives changes of task, verifier, and base model. A frozen model need not be a fixed agent: harness selection turns evaluation compute into durable capability.

cs.NE↗
arXiv2026

AutoPref: Automatic Discovery of Task-Specific Preference Objectives for Neural Combinatorial Optimization

Combinatorial optimization problems (COPs) underpin many real-world decisions, but their exponentially large search spaces make high-quality solutions costly to obtain. Neural combinatorial optimization (NCO) learns fast construction policies, typically with reinforcement learning (RL), while preference-based NCO improves sample efficiency by learning from relative solution quality. However, existing preference objectives combine two distinct design choices in manually specified, one-size-fits-all formulations: what learning signal to extract from each solution pair and how to weight each pair relative to the sampled set. We present AutoPref, the first LLM-guided framework for automated preference-objective discovery in NCO. AutoPref factorizes the objective into a pairwise loss program, which defines the learning signal, and a set-aware weighting program, which determines each pair's relative contribution. Their composition forms a unified programmatic objective space containing existing preference objectives as special cases. To make its search tractable, we introduce a staged conditional search strategy with behavioral gates that filter inadmissible programs before short-horizon training and evaluation. Across TSP, CVRP, FFSP, and JSSP, AutoPref consistently outperforms strong hand-designed baselines across problem scales, demonstrating the benefits and scalability of automated objective discovery for NCO.

cs.LG↗
arXiv2026

EvoPINN: Agentic Discovery of Executable Algorithms for Physics-Informed Neural Networks

Physics-informed neural networks (PINNs) have emerged as a powerful paradigm for solving partial differential equations (PDEs), yet their performance heavily relies on the manual, trial-and-error engineering of neural representations, loss formulations, and optimization dynamics. While Large Language Models (LLMs) offer a promising avenue for automated design, unconstrained code generation often yields mathematically invalid or numerically unstable solutions under strict scientific computing constraints. To bridge this gap, we propose \textbf{EvoPINN}, an agentic framework that reformulates PINN development from labor-intensive manual design into a rigorous, execution-grounded algorithm discovery problem. EvoPINN navigates a modular search space by decoupling neural representations from training programs, utilizing an LLM agent to iteratively propose memory-conditioned programmatic modifications. To ensure scientific validity, all candidates undergo strict structural verification and budget-matched PDE evaluation. Extensive experiments across diverse PDE regimes (oscillatory, elliptic, dissipative, and nonlinear transport) demonstrate that EvoPINN discovers PDE-specialized learning algorithms that significantly reduce relative $L_{2}$ error compared to baselines. Crucially, EvoPINN autonomously invented SLRC-PINN, a novel architecture whose performance gains persist under rigorous parameter-matched comparisons, establishing the viability of execution-grounded agents for discovering genuinely new scientific computing mechanisms.

cs.AI↗
arXiv2026

FAVA: Formal Authorization for Verified Agents with Evidence-Backed Permission Graphs

Large language model (LLM) agents autonomously interleave semantic reasoning with complex system operations. In these dynamic environments, static tool-level permissions are fundamentally insufficient; safe authorization is highly context-dependent and heavily reliant on evolving runtime states and data flows. We present FAVA (Formal Authorization for Verified Agents), a permission-carrying authorization framework for agent execution. FAVA utilizes an LLM-guided Permission Intermediate Representation (IR) to translate ambiguous natural-language tasks into structured constraints. A deterministic lowering pass then converts this IR into an evidence-backed permission graph that explicitly tracks data flows, dependencies, and contextual labels. To provide strict security guarantees, a Satisfiability Modulo Theories (SMT) authorizer mathematically verifies the current graph against security policies before any effectful action executes. A runtime gateway then enforces the solver's result, either authorizing the execution or intercepting it with a precise counterexample. We evaluate FAVA across OpenAgentSafety, OctoBench, and ActPlane scenarios. Our evaluation demonstrates that FAVA achieves a 90.5% Decision Compliance Rate (DCR) over the aggregate dataset, successfully intercepting dynamic violating traces in the evaluated trace-conditioned scenarios.

cs.CR↗
arXiv2026

ViSAGE: Constructing Self-Correcting Memories for Long-Form Video Understanding

Multimodal agents operating in long-horizon environments must build and continually update multimedia memories to support entity-consistent, temporally grounded reasoning. However, existing agentic memory approaches often discard fine-grained dentity cues under aggressive compression and segment-wise processing. They also rely heavily on vector similarity retrieval, which can surface semantically related yet identity-mismatched evidence, leading to entity confusion, error propagation, and hallucinated answers. We propose ViSAGE, a multimodal agentic memory framework that constructs self-correcting, entity-centric memories. Specifically, ViSAGE anchors entity identity via cross-modal binding over long temporal ranges. It then applies bidirectional memory refinement to propagate delayed identity evidence, retroactively unifying historical records and improving future reasoning. We also introduce multi-agent cross-verification to assess retrieved evidence under an identity-evidence alignment onstraint, enabling abstention instead of unsupported answers when evidence is missing. Extensive results demonstrate that ViSAGE consistently outperforms the strongest baseline, achieving 5.9% higher accuracy.

cs.AI↗
arXiv2026

When Good Equations Get Bad Scores: Improving Symbolic Regression Through Better Parameter Optimization

Symbolic Regression (SR) plays a central role in scientific knowledge discovery by distilling mathematical equations from observational data. Most existing SR methods function within a bi-level optimization framework: an outer loop that searches for the discrete equation structure, and an inner loop that optimizes the continuous parameters of that structure. Crucially, parameter-fitting quality directly determines a structure's score and thus the outer-loop search. However, nonlinear operators make the inner loop highly non-convex, and budget-driven reliance on fast local solvers (e.g., BFGS) often yields poor local minima and underestimated scores for correct structures. This ``Good Structure, Bad Score'' phenomenon becomes a key bottleneck, degrading efficiency and misguiding the search away from the true equation. To resolve this, we propose SAGE-Fit (Structure-Aware and Semantics-Guided Evaluator for Symbolic Regression), an SR-native fitting framework that exploits the dual native priors of symbolic expressions. By capitalizing on the structural and semantic priors unique to SR, we design tailored modules for each property, thereby effectively mitigating this optimization bottleneck. Extensive experiments demonstrate that our approach, as a plug-and-play module, significantly enhances evaluation fidelity and universally improves the performance of various SR systems.

cs.LG↗