arXiv subjects

Jiayi Liu

Publications and source records attributed to Jiayi Liu.

At least 19 recordsLinked to original sources

arXiv2026

Self-Evolving Agents through Privileged Retrospective Distillation

Self-evolving agents aim to improve from their own experience rather than remain fixed after deployment, but naturally collected trajectories often lack ground-truth solutions or verifier-derived rewards. We introduce Retrospective Self-Distillation (RSD), a multi-round framework that turns such experience into dense supervision using privileged retrospective context. For each on-policy trajectory, the same underlying model acts as both student and teacher: the teacher re-scores the student's realized actions while conditioning on cross-task precedents and lessons, together with turn-level reflection and future hindsight from the completed trajectory. The resulting teacher distribution is distilled back into the policy through on-policy self-distillation. Across rounds, the experience bank grows and the retrospective context is refreshed, allowing the model, its data, and its supervision to co-evolve. We further introduce frontier-guided context-policy evolution, where a stronger frontier model proposes constrained edits to how privileged context is constructed. Across ALFWorld, ScienceWorld, and SearchQA with Qwen3-8B and Qwen3-14B, RSD improves performance by up to 59.7% relative to the base model, while frontier-guided context-policy evolution adds up to a further 44.7% relative improvement over a fixed context policy. These results show that naturally accumulated experience can be converted into persistent model capability without ground-truth trajectories or verifier-derived rewards for the core policy update.

cs.LG↗

arXiv2026

Understanding and Mitigating Training Instability in Large-Scale Recommender Systems: A Cascade Perspective

Training large-scale recommender systems on dynamic industrial platforms, such as Short-Form Video (SFV) services with billions of daily interactions, faces critical stability challenges. Rapid distribution shifts in user data and increasing model complexity frequently trigger a divergence cascade, typically initiating with gradient explosion, propagating to weight explosion, and culminating in loss explosion. This G-W-L cascade severely impedes model convergence and performance. This paper provides a systematic analysis of this G-W-L cascade within an industrial SFV recommender context. We then propose and evaluate a structured mitigation framework where interventions are explicitly mapped to counteract specific stages of the cascade: gradient (G) control, weight/activation (W) stabilization, and loss (L) shaping. Our contribution provides a structured framework for understanding and tackling training divergence, offering practical insights for deploying robust, continuously updated recommenders in large-scale, non-stationary settings. We empirically validate the effectiveness of these strategies on a major industrial SFV recommendation system.

cs.IR↗

arXiv2026

EvoHarness-RL: Learning Runtime Harness Coordination for Self-Evolving Agents

Long-horizon LLM agents increasingly rely on external execution support to maintain state, track progress, recover from failures, and reuse experience across extended interactions. Yet existing harnesses and their use are often tailored to environments and controlled through prompts, heuristics, or system-specific rules, making agent and harness coordination difficult to jointly optimize. We introduce EvoHarness-RL, a unified framework that separates environment-specific harness implementations from a shared policy-facing interface. EvoHarness-RL organizes external support into a Belief, Progress, and Experience (BPE) workspace and exposes four compact harness actions for accessing and updating this state. We first instantiate BPE as an inference-time scaffold and then make harness coordination learnable through supervised initialization followed by cost-aware GRPO. Across heterogeneous long-horizon tasks, EvoHarness-Base improves the average success rate of frontier models by 10.0 percentage points, while EvoHarness-RL outperforms the strongest open-source baseline by 8.5 percentage points, together with higher RL rollout efficiency and stronger generalization to unseen tasks. Our analyses show that training gradually shifts agents from frequent scaffold use toward selective, environment-dependent harness access as the policy becomes more capable, while the external workspace continues to evolve and refine itself to better support task execution and generalization. Together, these results show that long-horizon agents benefit not only from external scaffolding itself, but also from learning how and when to coordinate with external support as part of a cost-aware policy.

cs.LG↗

arXiv2026

Residual Visual Credit Optimization: Conserved Evidence Routing for Multimodal Reinforcement Learning

Reinforcement learning with verifiable rewards scales multimodal reasoning, but an outcome reward says how much a trajectory is worth, not how that value should be spread over the decisions that produced it. We introduce Residual Visual Credit Optimization (RVCO), which treats token credit as a conserved routing problem. A controlled visual intervention yields a per-token evidence response; robust within-trajectory coordinates remove incidental scale; and a budgeted entropic router distributes a fixed amount of sequence utility according to perceptual dependence. A residual support path guarantees positive credit at every valid position, and an analytic correction restores the prescribed credit mass exactly. The resulting field is selective, bounded, full-support, and invariant to response-local score shifts, and recovers hard token selection as a limiting case. Across four model families and seven reasoning benchmarks, RVCO improves accuracy over strong RLVR baselines while maintaining late-stage optimization stability, corruption robustness, and competitive training cost. Rewards, rollouts, and the group-relative advantage estimator are unchanged; only the geometry of token-level credit differs.

cs.LG↗

arXiv2026

GISTBench: Evaluating LLM User Understanding via Evidence-Based Interest Verification

We introduce GISTBench, a benchmark for evaluating Large Language Models' (LLMs) ability to understand users from their interaction histories in recommendation systems. Unlike traditional RecSys benchmarks that focus on item prediction accuracy, our benchmark evaluates how well LLMs can extract and verify user interests from engagement data. We propose two novel metric families: Interest Groundedness (IG), decomposed into precision and recall components to separately penalize hallucinated interest categories and reward coverage, and Interest Specificity (IS), which assesses the distinctiveness of verified LLM-predicted user profiles. We release a synthetic dataset constructed on real user interactions on a global short-form video platform. Our dataset contains both implicit and explicit engagement signals and rich textual descriptions. We validate our dataset fidelity against user surveys, and evaluate eight open-weight LLMs spanning 7B to 235B parameters, together with three proprietary frontier models (GPT-5, Claude 4.6, and Gemini 3.5 Flash). Our findings reveal performance bottlenecks in current LLMs, particularly their limited ability to accurately count and attribute engagement signals across heterogeneous interaction types.

cs.AI↗

arXiv2026

Neuro-Symbolic Predicate Learning for Semantic Safe Robot Control

As robots are increasingly deployed in everyday environments, ensuring their safety has become a central challenge. Existing methods often encode safety requirements as opaque mathematical/logical formulations or dense cost functions. While effective in specific tasks, they remain difficult to interpret, tightly coupled to individual tasks, and offer limited insight into why a robot action is considered safe or unsafe. To address this limitation, we propose ``Neuro-Symbolic Predicate Learning for Semantic Safe Robot Control'' (NEUPRO), which leverages a differentiable reasoner that can learn reusable safety representations from human-specified safety knowledge. NEUPRO allows practitioners to express task-related safety requirements as transparent symbolic rules, while enabling gradients to propagate through these rules to a feature extractor that maps raw observations to safety-relevant concepts. As a result, the learned feature extractor is (softly) grounded in human-understandable semantics, supports transparent constraint evaluation, and is transferable across tasks. By coupling interpretability with differentiability, NEUPRO moves beyond opaque cost design toward reusable safety reasoning. To evaluate NEUPRO's capability, we collect and release REASON, the first real robot benchmark dataset for interpretable robot safety specification. Experiments on REASON show that NEUPRO learns safety-critical features that generalize across tasks, mitigate the interpretability limitations of conventional black-box cost formulations, and provide explicit explanations of safety violation.

cs.RO↗

arXiv2026

Gauss What You Need: Compact Gaussian Splatting Across Scene Scales

3D Gaussian Splatting reconstructs a scene as a collection of Gaussian primitives from a set of posed photographs called the capture. The number of primitives used to represent the scene affects reconstruction quality, storage, and rendering cost. How to select this number automatically across capture scales remains unresolved: configurations effective on standard benchmarks can leave larger captures with too few Gaussians to reconstruct fine details. We observe that the surface to represent, given by the capture's extent and resolution, is known before training, whereas its content complexity becomes apparent during training, through the reconstruction quality on the training views. We introduce TangoGS, which combines capture-derived model sizing with training-based adaptation: the capture determines the scale of the model, and training feedback determines its final size within that scale. Before training, TangoGS derives a learning allowance for model growth from the capture's total pixels after discounting views that re-observe the same scene points. During training, reconstruction quality guides how many Gaussians to add and remove. On 13 standard benchmark scenes, TangoGS matches the mean PSNR of the best-performing evaluated baseline, LeGS, with $48\%$ fewer Gaussians. On eight large captures, the same configuration automatically scales to larger models when necessary, achieving the highest mean PSNR among evaluated methods: $0.54$ dB above the runner-up with $2.3\times$ as many Gaussians. Together, capture-derived learning allowances and training-quality guided density control enable a state-of-the-art quality--size compromise across scene scales without retuning.

cs.CV↗

arXiv2026

Magnetoelectric Phase Transition and Axion Dynamics

Magnetoelectric phase transitions have been experimentally studied, but no macroscopic theory has been proposed to explain their dynamical origin. In this work, we assume that the axion quasiparticle with frequency undergoes a condensation like process. We show that these magnetoelectric phase transitions can be described within a Ginzbur Landau framework by introducing a coupled dynamic parameter, the axion angle, which is proportional to the magnetoelectric coeffcient. We derive relations between the static axion angle, the axion frequency, and the phase transition temperature for different magnetoelectric materials, respectively, and compare these calculations with existing experimental results. We also connect the artificially designed Dzyaloshinskii Moriya interaction with the axion condensate like process, so that the relation between the static axion angle and the experimentally measured frequency shift can be obtained.

cond-mat.other↗

arXiv2026

Decoupling Disaggregated Memory Optimizations from Indexing: A Compiler-Runtime Approach

Disaggregated memory (DM) decouples compute and memory into independently scalable pools, connected over a slower interconnect rather than a local bus. This decoupling is exactly what makes DM attractive--but it also means that every index must now reason explicitly about remote-memory access and its associated optimizations. State-of-the-art index designs respond to this by embedding remote-memory logic and optimizations directly into their core data structures and concurrency control mechanisms. Consequently, an optimization tuned for one index cannot be lifted and reused in another, and even the same index cannot be ported to a different DM architecture without a fresh round of redesign. This escalating, per-index, per-platform engineering burden is unsustainable as hardware and index requirements evolve. In this paper, we present Nox, a compiler-runtime framework that breaks this coupling by taking an unmodified, concurrent index as input and automatically generating its disaggregated-memory counterpart, without touching the original index logic. A compiler layer rewrites the index's LLVM IR to expose allocation, address, and pointer-dependency information that a centralized runtime uses to drive caching and address translation. Empirically, Nox-generated B+-trees, hash tables, and skip lists scale robustly on real RDMA and CXL hardware across every workload tested. They can also outperform some specialized, hand-crafted indexes and match others, especially on workloads that are closer to real-world ones. These results show that today's fastest disaggregated-memory optimizations need not stay locked inside monolithic, hand-crafted code--a compiler-runtime stack can generalize them while preserving the scalability of proven index implementations, without sacrificing it for portability.

cs.DB↗

arXiv2026

CORAL: An LLM-Native Harness for Production Recommender Systems

Production recommender systems shape what billions of people see, and sustaining their performance requires continual optimization: as content, user behavior, and upstream models shift, the choices governing retrieval, ranking, and serving must be revisited. Traditionally, human engineers test such changes through online experiments--a slow, reactive process limited by engineering effort, leaving parts of the system unrevised as conditions change. Although large language models have been applied to ranking, user modeling, and offline model development, few systems place an agent in a continual closed loop that acts on a live recommender and learns from the measured effects of its decisions. We present CORAL (Constraint-Optimized Recommender via an Agentic Loop), an LLM-native harness that closes this loop: each cycle, the agent observes operating signals, reasons over a memory of past decisions and outcomes, and invokes tools--including a numerical optimizer that keeps changes within a fixed operating budget--to reconfigure the recommender, with measured outcomes informing the next cycle. We formulate this as a partially observed, non-stationary, constrained optimization problem in which the policy improves in context, without parameter updates, from its prior actions. Across two large-scale social platforms, evaluated with A/B experiments, the same harness improves engagement at no additional serving cost on one and reduces serving cost without degrading engagement on the other, spanning the engagement-efficiency frontier. Performance improves as the loop iterates, suggesting that a single agentic loop can automate continual optimization work traditionally performed by human algorithm engineers under explicit guardrails.

cs.CL↗

arXiv2026

LD4WAM: Learning Latent Dynamics from Human Videos for World Action Models

Human video is playing an increasingly central role in training World Action Models (WAMs), owing to its diversity and low collection cost relative to teleoperated robot data. However, most WAMs learn from such video only by predicting pixel-level future frames, giving dynamics that are not directly actionable, whereas motion retargeting recovers directly actionable actions but leaves a large visual gap across embodiments. We therefore propose motion-aligned latent dynamics as an embodiment-agnostic representation to bridge video priors and low-level actions. We further present LD4WAM, which pairs a Latent Dynamics Model trained with semantic reconstruction and real motion alignment with a World Dynamics Action Model built as a mixture-of-transformers (MoT), which preserves full future-video generation and uses learnable queries to distill these latent dynamics from generated futures for action conditioning. Pretrained on our curated unified dataset of over 5{,}000 hours of human and robot data, LD4WAM performs strongly in RoboTwin simulation and on real robots equipped with both grippers and dexterous hands, while generalizing well to unseen objects and backgrounds.

cs.RO↗

arXiv2026

Building An Integrated Vector Database System in PostgreSQL

This paper presents PostgreSQL-V 2.0, a scalable integrated vector database system inside PostgreSQL. Existing PostgreSQL-based vector search systems such as pgvector embed vector indexes into PostgreSQL's page-oriented storage engine, incurring significant overhead that leads to a huge performance gap with specialized vector databases. In our earlier work, we introduced PostgreSQL-V 1.0, which addresses this issue by separating vector index structures from PostgreSQL's storage engine, enabling vector search performance close to that of native vector index libraries while preserving SQL compatibility. However, we find that PostgreSQL-V 1.0 has three limitations that matter for real-world workloads: it only supports a single connection (without concurrency), recovery time grows with index size, and physical replication is unsupported. We further present PostgreSQL-V 2.0, which closes all three gaps. PostgreSQL-V 2.0's concurrency support enables fully concurrent vector searches and updates across PostgreSQL's multi-process backends, delivering up to 36.4x the throughput of PostgreSQL-V 1.0 while serving 32 concurrent clients. PostgreSQL-V 2.0's fast crash recovery keeps cost independent of total index size, remaining near 20 ms while PostgreSQL-V 1.0's grows into seconds-scale. PostgreSQL-V 2.0's physical replication support extends physical replication to the decoupled index, preserving index consistency on standbys without burdening the primary node. Together, these advances make PostgreSQL-V 2.0 a fully concurrent, crash-resilient, and replication-ready vector database inside PostgreSQL.

cs.DB↗

arXiv2026

Axion Dark Matter Modulated Spin Wave Interferometry

We propose a novel asymmetric spin wave interferometer to detect ultralight axion dark matter. The axion modulates spin-wave properties via a weak effective magnetic field in ferromagnets. The interferometer splits a spin-wave source into two paths of different lengths and sets them to interfere destructively. The system then converts the axion-induced phase shift into a measurable magnetization oscillation that can radiate electromagnetic waves and generate electrical signals via Faraday induction. The signal-to-noise ratios have been evaluated for three detection schemes: the linear amplifier, the single-photon detector, and the electrical signal detection approach, accounting for both magnetization fluctuation and thermal noise. The accessible axion mass range is approximately $10^{-8}$ eV to $10^{-6}$ eV, set by the spin wave propagation length and the relaxation time.

hep-ph↗

arXiv2026

Color Fundus Photography Analysis: Co-evolution of Data, Preprocessing, and Modeling toward Multimodal AI

Color Fundus Photography (CFP) is a primary non-invasive imaging modality for large-scale screening of ophthalmic and systemic diseases. Existing surveys mainly summarize task-specific algorithms, datasets, or preprocessing techniques independently, lacking a unified perspective on their co-evolution with modern artificial intelligence. This review provides an integrated overview of CFP AI through the interplay of dataset evolution, preprocessing paradigms, and modeling frameworks. We show that CFP datasets have evolved from small single-center collections with task-specific labels to large multi-center resources featuring multimodal pairings and longitudinal clinical records. Preprocessing has progressed from conventional image enhancement to neural data-engineering pipelines, hardware-aware token optimization, and self-supervised imputation for incomplete electronic health records (EHRs). Meanwhile, modeling has advanced from convolutional neural networks (CNNs) to vision foundation models, state space models (SSMs), and multimodal expert architectures. At the multimodal frontier, CFP is increasingly integrated with EHRs and longitudinal patient information, enabling more comprehensive clinical reasoning beyond isolated image analysis. We conclude that future progress depends on the collaborative optimization of datasets, preprocessing, and multimodal modeling, providing a roadmap toward robust clinical deployment, improved cross-domain generalization, and resource-efficient edge intelligence.

cs.CV↗

arXiv2026

Efficient Analytic Uncertainty Quantification for Multi-Modal Regression

Efficient uncertainty quantification (UQ) is essential for trustworthy large-scale learning. Existing UQ methods for regression tasks mainly operate under the assumption that the conditional label marginal satisfies single-peak parametric models, e.g., Gaussians, where the negative log-likelihood function simplifies to the mean square error. However, such single-peak assumptions fail in regression tasks featuring multi-modal distributions. On the other hand, semi-parametric methods which achieve strong regression performance for multi-modal distributions often lack efficient quantification on their prediction variances. In this work, we extend UQ techniques based on Variational Bayesian Inference (VBI) to two widely used semi-parametric regression models that yield histogram-like reconstructions of the conditional label densities: Quantile Regression (QR) and Classification Restoration (CR). Our approach introduces a unified, distribution-agnostic framework that simultaneously achieves accurate estimation of complex conditional distributions and highly efficient UQ. Theoretically, our method is grounded in novel formulations of QR and CR within the VBI framework, yielding analytic Evidence Lower Bounds (ELBO) to streamline training and a closed-form or analytically approximated predictive density for efficient inference. Empirically, we evaluate our methods on three large-scale regression benchmarks with multi-modal label distributions. Our framework outperforms state-of-the-art multi-modal regression baselines, and even matches predictive performance of computationally expensive ensemble models. Furthermore, by leveraging epistemic uncertainty estimation, our approach enables highly data-efficient active learning strategies.

cs.LG↗

arXiv2026

SAGE-OPD: Selective Agent-Guided Intervention for Multi-Turn On-Policy Distillation

On-policy distillation (OPD) improves student models by training them on trajectories induced by their own policy, making it a promising approach for mitigating exposure bias in agent training. However, most OPD studies focus on single-turn settings, while realistic LLM agents interact with environments over multiple turns. In this regime, early errors can alter future observations and compound across the trajectory, and standard dense token-level OPD becomes brittle, as it may over-penalize semantically valid alternatives, reinforce local degeneracies such as repeated actions, and propagate unreliable teacher supervision on off-distribution histories. We propose SAGE-OPD, a verifier-free selective intervention framework specifically designed for multi-turn OPD. Instead of applying teacher supervision uniformly across all turns, SAGE-OPD first observes environment feedback and uses teacher judgment to decide whether each student response should be skipped or intervened on. To further address compounding errors, SAGE-OPD weights token-level distillation by teacher confidence, reducing the influence of uncertain teacher distributions on corrupted or ambiguous histories. Finally, SAGE-OPD applies loss normalization to preserve the overall loss scale of standard OPD while retaining selective turn-level weighting. Experiments on agent tasks show that SAGE-OPD consistently improves over baselines, achieving up to a 13.3% relative improvement in ALFWorld unseen success rate over standard OPD. Ablation studies further demonstrate that turn-level intervention, teacher confidence weighting, and loss normalization provide complementary benefits. Our results suggest that effective multi-turn OPD should remain on-policy, but teacher supervision should be selectively allocated to turns where intervention is necessary and reliable.

cs.CL↗

arXiv2026

OmniOPD: Logit-Free On-Policy Distillation via Speculative Verification

On-Policy Distillation (OPD) trains a student model on its own generative trajectories under dense token-level feedback from a stronger teacher, mitigating both the off-policy distribution shift of Supervised Fine-Tuning (SFT) and the sparse credit assignment of Reinforcement Learning (RL). However, standard OPD faces two coupled limitations. First, it requires direct access to the teacher's token-level logits, excluding a broad class of capable proprietary models from serving as teachers. Second, the token-level logit signal itself is brittle, depending on a narrow overlap of plausible next tokens between teacher and student, and prone to amplifying degenerate patterns such as repetition loops. In this paper, we introduce OmniOPD, a novel framework that addresses both limitations through a logit-free, chunk-level supervision signal. OmniOPD replaces deterministic logit matching with Monte Carlo rollouts that approximate the teacher's local preferences through a continuous semantic similarity metric over multi-token chunks, and concentrates this supervision via a peak-entropy scheduler that audits the student only at its high-uncertainty reasoning forks. A Dirichlet-Multinomial Bayesian prior and a base-model KL anchor further bound the variance of discrete sampling and prevent policy collapse across unaudited tokens. Across competitive benchmarks, OmniOPD surpasses the standard OPD approach by up to +28.64% on math, confirming that chunk-level semantic verification extracts a more reliable learning signal than token-level logit matching, whose high information density is offset by significant noise and brittleness. Furthermore, when paired with stronger black-box teachers such as Claude-4.5-Haiku and Gemini-2.5-Flash, OmniOPD achieves an additional +9.54% relative on math over its open-weight teacher counterpart, advancing the student past the performance of self-exploratory RL.

cs.LG↗

arXiv2026

DAG-MoE: From Simple Mixture to Structural Aggregation in Mixture-of-Experts

Mixture-of-Experts (MoE) models have become a leading approach for decoupling parameter count from computational cost in large language models, yet effectively scaling MoE performance remains a challenge. Prior work shows that fine-grained experts enlarge the space of expert combinations and improve flexibility, but they also impose substantial routing overhead, creating a new scalability bottleneck. In this paper, we explore a complementary axis for scaling -- how expert outputs are aggregated. We theoretically show that replacing the standard weighted-summation aggregation with structural aggregation expands the expert-combination space without altering the experts or router, and enables possible multi-step reasoning within a single MoE layer. To this end, we propose DAG-MoE, a sparse MoE framework that employs a lightweight module to automatically learn the optimal aggregation structure among the selected experts. Extensive experiments under standard language modeling settings show that DAG-MoE consistently improves performance in both pretraining and fine-tuning, surpassing traditional MoE baselines.

cs.AI↗