arXiv subjects

Tong Chen

Publications and source records attributed to Tong Chen.

At least 19 recordsLinked to original sources

arXiv2026

Aligning Autoresearch: Co-Improvement between Agents and Humans

Autoresearch agents can now attempt long-horizon AI research end-to-end, i.e., self-improvement. However, best results are still obtained by research agents interacting with human researchers, and improving AI's ability to perform this co-research will only reinforce this effect---a process termed co-improvement (Weston and Foerster, 2025). Further, aligning with human expectations is likely a safer path for research agents, especially given recent agent reward hacking phenomena (OpenAI, 2026b). In this work we address these issues in four ways: (i) we empirically study the role of humans in directing autoresearch agents, focusing on the domain of benchmark creation, and find humans still play a pivotal role particularly in the ideation phase; (ii) we also develop a benchmark, HackBench, that shows current autoresearch agents are not aligned and are unable to achieve hack localization sufficiently; (iii) we present a method to train agents for improved human interaction by using human simulators as a proxy during reinforcement learning (RL); and finally (iv) we develop a method to improve model text generation, and hence agent communication, by RL with rubrics aligned with human experts. Together, our training and evaluation recipes provide a path toward more capable and aligned (co-)research agents.

cs.AI↗

arXiv2026

Learning Projection-Aware 360-Degree Image Rectification via Dual-Projection Fusion

Panoramic cameras provide a 360° field of view and are widely used in panoramic vision, immersive visual computing, and robotic perception. However, changes in camera orientation can produce non-upright panoramas, introducing geometric variations that can complicate downstream visual analysis. Existing vision-based rectification methods usually operate within a single projection domain, limiting their ability to jointly exploit local geometric structures and global contextual information. To address this, we formulate 360° image rectification as a projection-aware representation learning problem and propose a dual-projection framework for upright panoramic rectification. A convolutional neural network branch captures local geometric structures from equirectangular projection (ERP) inputs, while a vision transformer branch models global contextual cues from cubemap projections. Cross-projection feature transformation and multi-level feature fusion enable effective interaction between these complementary representations. The learned representation supports collaborative inclination estimation and upright panorama generation, with the two tasks providing complementary geometric and appearance supervision. Experiments on SUN360 and M3D show consistent improvements over existing methods, achieving accuracies within a 1° error threshold of 65.9% and 85.2% and Fréchet Inception Distance scores of 5.87 and 3.26, respectively. Ablation studies verify the contributions of dual-projection representation, cross-projection feature transformation and fusion, and collaborative multi-task learning. The proposed framework provides a projection-aware visual computing approach for panoramic rectification. Code, pretrained models, and training/testing scripts are available at https://github.com/YuhaoShine/DualProjectionFusion.

cs.CV↗

arXiv2026

pCoMole: Pareto-Constrained Molecule Editing with Discrete Flows

Biomolecular therapeutics often start from known sequences and require targeted editing to improve multiple properties while satisfying hard biochemical and manufacturability constraints. However, existing generative methods do not jointly support multi-objective optimization, hard feasibility, and sequence editing in discrete, variable-length biological spaces. In this work, we introduce Pareto-Constrained Molecule Editing (pCoMole), a framework built on discrete flow matching that steers a pre-trained Edit Flow toward user-specified preferences while enforcing terminal feasibility. pCoMole defines a feasibility-gated terminal distribution using an augmented Tchebycheff utility and realizes the resulting preference tilt through a Doob-h transform of the underlying edit process. To make this construction practical, we approximate the required harmonic function using short Monte Carlo rollouts over candidate edits, yielding an efficient guided editor with provable preference consistency. We validate pCoMole by shrinking GFP while retaining fluorescence-related properties, shortening diverse Cas9 orthologs while preserving PAM specificity, and compressing peptide binders into short peptidomimetics that optimize seven drug-related properties under hard constraints. In wet lab testing, two 229-residue pCoMole-designed eGFP variants retained clear green fluorescence in BL21 cells after 10 deletions, with either one or two substitutions. Together, pCoMole enables constraint-aware, Pareto-aligned editing of biomolecular sequences in discrete, variable-length spaces.

cs.LG↗

arXiv2026

Evaluation of OpenAI o1: Opportunities and Challenges of AGI

This comprehensive study evaluates the performance of OpenAI's o1-preview large language model across a diverse array of complex reasoning tasks, spanning multiple domains, including computer science, mathematics, natural sciences, medicine, linguistics, and social sciences. Through rigorous testing, o1-preview demonstrated remarkable capabilities, often achieving human-level or superior performance in areas ranging from coding challenges to scientific reasoning and from language processing to creative problem-solving. Key findings include: -83.3% success rate in solving complex competitive programming problems, surpassing many human experts. -Superior ability in generating coherent and accurate radiology reports, outperforming other evaluated models. -100% accuracy in high school-level mathematical reasoning tasks, providing detailed step-by-step solutions. -Advanced natural language inference capabilities across general and specialized domains like medicine. -Impressive performance in chip design tasks, outperforming specialized models in areas such as EDA script generation and bug analysis. -Remarkable proficiency in anthropology and geology, demonstrating deep understanding and reasoning in these specialized fields. -Strong capabilities in quantitative investing. O1 has comprehensive financial knowledge and statistical modeling skills. -Effective performance in social media analysis, including sentiment analysis and emotion recognition. The model excelled particularly in tasks requiring intricate reasoning and knowledge integration across various fields. While some limitations were observed, including occasional errors on simpler problems and challenges with certain highly specialized concepts, the overall results indicate significant progress towards artificial general intelligence.

cs.CL↗

arXiv2026

A Discrepancy-Based Perspective on Dataset Condensation

Given a dataset of finitely many elements $\mathcal{T} = \{\mathbf{x}_i\}_{i = 1}^N$, the goal of dataset condensation (DC) is to construct a synthetic dataset $\mathcal{S} = \{\tilde{\mathbf{x}}_j\}_{j = 1}^M$ which is significantly smaller ($M \ll N$) such that a model trained from scratch on $\mathcal{S}$ achieves comparable or even superior generalization performance to a model trained on $\mathcal{T}$. Recent advances in DC reveal a close connection to the problem of approximating the data distribution represented by $\mathcal{T}$ with a reduced set of points. In this work, we present a unified framework that encompasses existing DC methods and extend the task-specific notion of DC to a more general and formal definition using notions of discrepancy, which quantify the distance between probability distribution in different regimes. Our framework broadens the objective of DC beyond generalization, accommodating additional objectives such as robustness, privacy, and other desirable properties.

cs.LG↗

arXiv2026

RepoAtlas: Guiding Coding Agents via Evolving Multimodal Repository Views

Large language model (LLM)-powered coding agents have made rapid progress in automating software engineering tasks, yet repository-level issue resolution remains challenging. Beyond generating a plausible patch, an agent must localize relevant code across interdependent files and maintain repository context that is both sufficient and focused. Code graphs expose non-local relations, but linear text interfaces obscure their topology; rendering the full repository graph yields visual representations that are too dense to perceive reliably, whereas a one-shot local view becomes stale as exploration proceeds. We present \textbf{RepoAtlas}, a training-free module that maintains evolving multimodal repository views through a \emph{select--project--refresh} loop over a repository code graph. RepoAtlas combines evidence from the issue with the agent's current exploration state to select a task-relevant region under a fixed budget, projects the selected structure into complementary visual and textual representations, and refreshes the view when changes in the exploration state render it outdated. We evaluate RepoAtlas on SWE-bench Verified, where it improves the resolve rate by 2.4 points while reducing input tokens and model calls by 5.8\% and 7.8\% on average, relative to the strongest multimodal graph baseline, with consistent gains across three models of different families and scales.

cs.SE↗

arXiv2026

SinoGlyphBench: A Diagnostic Benchmark for Chinese Glyph-Level Obfuscation in Language-Model Moderation

Glyph-level obfuscation can leave harmful Chinese content readable to humans while degrading automated moderation. We introduce SinoGlyphBench, a diagnostic benchmark that identifies label-critical semantic anchors and creates matched original and glyph-obfuscated inputs in text and image modalities. By perturbing anchors, background context, or both, this design distinguishes corruption of moderation-relevant evidence from general surface variation. Across 176,916 paired evaluations of 12 LLMs and MLLMs, obfuscation increases harmful false-negative and false-positive rates by 6.1 and 4.7 percentage points, respectively, and reduces four-way accuracy by 5.0 points. Models retain 75.7% of the decisions that were correct on the matched original inputs. Full-scope perturbations cause the largest degradation, anchor-only perturbations are more damaging than background-only perturbations, and cross-script substitution is particularly difficult in the text modality. Analysis of structured outputs identifies observable mismatches in visible-form reading, intended-message recovery, and final safety judgment. The evaluated models, therefore, remain brittle to Chinese content written with non-canonical glyphs. Resources are available at https://github.com/fengshun124/SinoGlyphBench.

cs.CL↗

arXiv2026

Empowering Compact LLMs with Fusion of Layer-wise Exits for Recommendation

Large language model-based recommender systems (LLM-RSs) have demonstrated remarkable capabilities, but are computationally unsustainable for many real-world applications. Compact LLMs offer a practical alternative, yet their reduced capacity often requires reasoning or knowledge distillation methods that increase latency or depend on larger models. Combined with autoregressive generation, these approaches face severe scalability bottlenecks. In contrast, discriminative LLM-RSs enable efficient full-corpus ranking through embedding similarity, but compact backbones remain limited in expressiveness and structural adaptivity. We propose the Fusion of Layer-wise Exits for Sequential Recommendation (FLEXRec), a discriminative framework that enhances compact LLMs while retaining scalable full-corpus ranking. FLEXRec inserts prediction heads (i.e., exits) at multiple transformer layers and adaptively fuses their score distributions. An adaptive continuous router (AC-Router) dynamically selects both the number and identity of exits for each user sequence, while a novel target-k hinge loss regulates routing sparsity. Experiments on three real-world datasets with Qwen 3 1.7B and Llama 3.2 3B show that FLEXRec achieves state-of-the-art accuracy among competing methods while remaining highly efficient. Code: https://github.com/xurong-liang/FLEXRec

cs.IR↗

arXiv2026

Security and Privacy in Large-Model-Driven Embodied Agents: Attacks, Defenses, and Future Directions

Large-model-driven embodied agents integrate foundation models with perception, reasoning, planning, and physical action, extending conventional model-level risks into embodied closed loops. Existing studies on their security and privacy remain fragmented across different system components and operational stages, making it difficult to understand how risks arise, propagate, and ultimately affect physical behavior or sensitive information. This survey presents a lifecycle-based analysis of security and privacy in large-model-driven embodied agents. We organize existing research into five stages: model construction and supply chain, multimodal input and interaction, semantic reasoning and task planning, action execution and physical feedback, and long-term deployment. Within this lifecycle, we systematically review representative attacks, defenses, and evaluation methods. Our analysis shows that attack entry, consequence realization, and defense intervention often occur at different stages of the embodied closed loop. It further reveals substantial gaps in end-to-end protection, real-world evaluation, and long-term privacy governance. This survey provides a unified perspective for understanding current progress and identifying critical directions for securing large-model-driven embodied agents.

cs.CR↗

arXiv2026

QA-Merging: Query-Adaptive Reasoning via Layer Selective Model Merging

Recent large reasoning models (LRMs) have achieved strong performance on complex reasoning tasks by generating a long chain-of-thought (Long-CoT). However, such lengthy reasoning is often unnecessary for simple queries, leading to additional computation and latency. Existing approaches to adaptive reasoning typically rely on retraining the model or designing sophisticated prompting, which are either prohibitively expensive or highly sensitive to the prompt formulation. Model merging provides a more balanced alternative for adaptive reasoning by avoiding expensive training and integrating Long-CoT and Short-CoT behaviors. However, existing merging methods are often static and input-agnostic, or rely on costly all-layer calibration, which limits their effectiveness for query-adaptive reasoning. To tackle these challenges, we propose Query-adaptive Layer Selective Merging (QA-Merging), an activation-based merging framework that integrates a Long-CoT model and a Short-CoT model to obtain a query-adaptive reasoner without training from scratch or requiring large-scale additional data. QA-Merging first constructs a small pattern-labeled calibration set that assigns each query an appropriate reasoning pattern. Motivated by our empirical analysis that Long-CoT and Short-CoT behaviors diverge unevenly across transformer layers, QA-Merging identifies layers with high reasoning pattern divergence and calibrates only these layers through feature alignment and contrastive shaping, while applying closed-form hidden-state correction to the remaining layers. Experiments on seven widely used reasoning benchmarks across two model scales demonstrate that QA-Merging reduces inference cost and maintains strong performance.

cs.CL↗

arXiv2026

Towards Efficient Reasoning in LLM-Based Recommender Systems via Model Merging

Large language model-based recommender systems are increasingly adopting slow-thinking models that generate step-by-step reasoning before making predictions, often achieving higher accuracy than fast-thinking models that predict directly. However, their reasoning traces are often unnecessarily verbose, increasing inference costs without commensurate accuracy gains. Existing training-based approaches to reasoning compression often incur substantial adaptation costs, while inference-time methods are brittle and difficult to scale. These limitations motivate model merging as a promising training-free direction for transferring specialised behaviours between models in a shared parameter space. In particular, merging a slow-thinking model with a fast-thinking counterpart provides a natural mechanism for balancing recommendation accuracy and reasoning conciseness. To this end, we propose, to our knowledge, the first model merging framework for reasoning compression in recommender systems. Unlike conventional merging methods that apply uniform merge coefficients across model components, our method performs fine-grained merging at the level of individual attention heads, capturing heterogeneous patterns in recommendation reasoning. Each attention head is assigned a distinct merge coefficient according to its contribution to critical reasoning evidence and its sensitivity to parameter change, enabling selective injection of the concise behaviour of the fast-thinking model into the slow-thinking model and reducing reasoning verbosity without compromising recommendation quality. Experiments on three benchmark datasets show that our method reduces reasoning length by up to 24.3% while outperforming competitive model merging baselines in maintaining recommendation accuracy. The code is available at https://github.com/linhledieu/REAM.

cs.IR↗

arXiv2026

Surg-UniWorld: A Unified Surgical World Model with Multimodal Control Experts

Controllable surgical world models can provide a generative foundation for surgical artificial intelligence and simulation by synthesizing realistic instrument--tissue interactions. However, existing methods lack a unified multimodal control paradigm, while direct fusion of heterogeneous visual conditions often causes anatomical distortion, instrument appearance drift, and temporally inconsistent interactions. In this work, we propose {Surg-UniWorld}, a unified surgical world model with multimodal control experts. Surg-UniWorld first constructs a {Hierarchical Surgical Anchor} from first-frame appearance and hierarchical semantic masks to preserve persistent scene identity, anatomical organization, and interaction boundaries. {Anchor-Relative Modality Experts} then interpret edge, depth, and optical-flow evidence relative to the shared anchor, capturing complementary boundary, geometric, and motion information. A {Multimodal Control Expert} further performs contribution-preserving stage-wise composition of the activated modality increments and generates control hints for the Wan2.2 video diffusion backbone. To support multimodal surgical world modeling, we further construct Cholec80-SurgWAM, a benchmark for controllable surgical video generation. Extensive experiments demonstrate that Surg-UniWorld consistently outperforms existing controllable video generation methods and surgical world-model baselines in generation quality, temporal consistency, and multimodal controllability.

cs.AI↗

arXiv2026

VaLiDRec: Variable-Length LLM-Aligned Semantic IDs for Generative Recommendation

Generative recommendation commonly represents items using fixed-length semantic identifiers (SIDs) constructed through clustering and quantization. However, these artificial codes may overcompress item semantics, remain misaligned with pretrained LLM vocabularies, and require costly autoregressive decoding. In light of this, we propose VaLiDRec, a generative recommendation framework based on variable-length, LLM-aligned semantic identifiers. VaLiDRec constructs SIDs directly from informative native LLM vocabulary tokens via token importance estimation, semantic-quality-aware pruning, and collision-aware refinement, allowing identifier lengths to adapt to item semantic complexity. To model user preferences, VaLiDRec incorporates graph-aware soft prompts and reformulates recommendation as token-set prediction with token-level item scoring, eliminating autoregressive SID generation and beam search. Experiments on four real-world datasets show that VaLiDRec consistently outperforms strong sequential and generative recommendation baselines across all evaluation metrics. It further achieves superior zero-shot item cold-start performance and 87.49$\times$ faster inference than LC-Rec. These results demonstrate that LLM-native variable-length semantic identifiers provide a more expressive and efficient paradigm for generative recommendation.

cs.IR↗

arXiv2026

ODYSSE: Episode-wise Policy Optimization for Personalized Agentic Reasoning

Agentic systems have rapidly advanced in their ability to interact with real-world environments, leverage external tools, and provide services for users. However, unlike natural-world tasks that assume well-defined instructions, human-centered scenarios are characterized by ambiguous requests that lead to large, open-ended solution spaces. Decoding users' personalized preferences is therefore essential for narrowing the candidate solution space. This introduces a new challenge, personalized agentic reasoning, which requires agents to jointly interact with both users and environments to deliver personalized services. In this paper, we present ODYSSE, a Reinforced Fine-Tuning (RFT) framework for personalized agentic reasoning. At its core, ODYSSE proposes Episode-wise GRPO (ESPO), a novel extension of Group Relative Policy Optimization (GRPO) designed to address long action horizons and strong cross-step dependencies in personalized agentic reasoning. Rather than optimizing individual steps independently, ESPO introduces an episode-level reward mechanism together with episodic advantage estimation, enabling upstream evidence to effectively guide downstream personalized decisions and allowing agents to progressively resolve ambiguous user requests across multiple interaction steps. We further propose an episodic batch sampler that groups actions from the same episode into unified training batches, facilitating coherent optimization under ESPO. We evaluate ODYSSE on realistic long-horizon personalized GUI reasoning tasks. Experimental results demonstrate that ODYSSE consistently outperforms both specialist and general-purpose LVLMs, highlighting its effectiveness for personalized agentic reasoning.

cs.AI↗

arXiv2026

Revealing the $χ_{\rm eff}$-$q$ Correlation among Coalescing Binary Black Holes and Tentative Evidence for AGN-driven Hierarchical Mergers

The origin of the correlation between the effective spins ($χ_{\rm eff}$) and mass ratios ($q$) of LIGO-Virgo-KAGRA's binary black holes (BBHs) is still an open question. Motivated by the recent identification of two subpopulations of the BBHs, in this work we investigate the potential $χ_{\rm eff}-q$ correlation for each subpopulation. Surprisingly, the $χ_{\rm eff}$-$q$ correlation {either significantly weakens or disappears} for the low-mass subpopulation if we introduce a second $χ_{\rm eff}$ distribution for the high-mass subpopulation, which likely originates from hierarchical mergers. {This suggests that the $χ_{\rm eff}$-$q$ correlation in the overall population can be explained by the superposition of two distinct subpopulations.} {We find Bayesian evidence strongly favoring two separate $χ_{\rm eff}$ distributions over a single mass-ratio-dependent distribution, with Bayes factors $\ln\mathcal{B}>4.2$.} The first subpopulation has a narrow $χ_{\rm eff}$ distribution peaking at $\sim0.05$, whose primary-mass function {showing a rapid decline beyond} $\sim 40M_{\odot}$, in agreement with first-generation BBHs. The second $χ_{\rm eff}$ distribution is broad and peaks at $μ_{χ,2} \sim 0.4$, aligning with predictions for hierarchical mergers in active galactic nucleus (AGN) disks. {However, we cannot exclude negative $χ_{\rm eff}$values in the second subpopulation, suggesting hierarchical mergers might occur both in AGN disks and stellar clusters. Furthermore, the inferred second $χ_{\rm eff}$ distribution might alternatively arise from other formation channels, such as stable mass transfer or chemically homogeneous evolution, if not interpreted as hierarchical mergers.}

astro-ph.HE↗

arXiv2026

Field-induced first order transitions and phase coexistence in the Kitaev quantum spin liquid candidate BaCo2(AsO4)2

BaCo2(AsO4)2 (BCAO) is an insulating Kitaev quantum spin liquid candidate with a rich low-temperature phase diagram. Below 5 K, it exhibits double-zigzag magnetic order. Upon application of an in-plane magnetic field, the magnetic structure first transforms into an up-up-down (UUD) state near 0.12 T and then enters a fully spin-polarized ferromagnetic (FM) state near 0.5 T. In the narrow field regime close to the polarized phase, a finite residual thermal conductivity has been reported, suggesting a possible field-induced quantum spin liquid phase. Using neutron scattering, we show that the field-induced double-zigzag-to-UUD transition near 0.15 T is accompanied by the emergence of a cluster of localized UUD spin excitations that coexist with conventional spin waves. Upon further increasing field, BCAO undergoes a first-order UUD-to-FM transition with coexistence of UUD and FM phases, accompanied by a sharp response in the spin Seebeck coefficient. These results do not support a quantum spin liquid scenario near the UUD-to-FM critical field. Instead, modeling indicates that the broad excitations arise from bound spin-flip pairs, while a low-lying dispersive branch near the FM phase boundary carries the same sign of magnetization as the FM order. These excitations naturally account for the observed sign of the spin Seebeck response and are likely relevant to the thermal conductivity near the critical field.

cond-mat.str-el↗

arXiv2026

SpeedyGS: Content-Aware 3D Gaussian Splatting Compression via Two-Stage Optimization

Recent progress in compressing large-scale 3D Gaussian Splatting (3DGS) data has substantially reduced storage footprint, network transmission bandwidth, and memory traffic to GPU caches before rendering. Yet decoding with advanced 3DGS codecs still takes seconds, making them unsuitable for interactive applications. To systematically address this challenge, we propose SpeedyGS, a Content-Aware 3DGS Compressor that separately optimizes the structural formation and statistical coding. First, in structural formation, we jointly optimize adaptive quantization and pruning under a unified rate-distortion objective, where the rate term is replaced by a lightweight rate proxy that estimates entropy coding cost of the next stage, thereby efficiently regulating Gaussian density and precision to yield a compact scene representation. Then, in the statistical coding phase, Gaussian geometry is converted into sparse octree tokens and subsequently undergoes multi-stage coding, while Gaussian attributes are serialized into a 1D token stream for entropy coding via a complexity-controllable local autoregressive model. SpeedyGS achieves a favorable balance among optimization efficiency, compression performance, decoding latency, and rendering speed. Compared to vanilla 3DGS, SpeedyGS achieves up to 160$\times$ model size reduction with negligible quality degradation across common datasets. Compared to state-of-the-art compression methods, it also offers significantly faster decoding and accelerates optimization by 9$\times$ on consumer-grade hardware. To further reduce decoding overhead, the statistical coding stage also supports channel-wise, fixed-length coding for Gaussian as a simpler alternative, enabling SpeedyGS to better adapt to the underlying application and reduce decoding latency to nearly zero.

eess.SP↗

arXiv2026

What Would You Click? Personalized Video Thumbnail Generation with Preference-aware Highlight Retrieval

Video thumbnails are a key factor for attracting user clicks on video platforms, and are increasingly supported by automation. However, existing thumbnail generation methods typically produce generic results shared across users, overlooking the diversity of individual preferences. We therefore introduce personalized video thumbnail generation, a novel task that aims to create thumbnails tailored to user-specific preferences. It is challenging in two aspects: (i) identifying visual anchors (i.e., key frames) from each video to guide the generation, which requires a balance between personalization and informativeness that existing highlight detection methods fail to achieve; and (ii) generating personalized thumbnails that are both visually coherent and faithful to the original video. As a response, we propose a two-stage framework that tightly couples preference-aware retrieval with controllable generation. In the first stage, a personalized highlight retriever captures fine-grained user-video interactions and incorporates video semantics through summarization, enabling the selection of diverse visual anchors aligned with both user preferences and video contexts. In the second stage, a VLM-guided diffusion pipeline transforms these anchors into thumbnails by extracting and injecting semantically grounded visual cues, improving personalization while preserving visual coherence and fidelity. Experiments on two public datasets show our method delivers state-of-the-art performance compared with both retrieval-based and generative baselines. A user study further demonstrates improved click preference, highlighting its effectiveness in enhancing user engagement. The code is available at https://github.com/hezy18/PVTG.

cs.MM↗