Model Editing for New Document Integration in Generative Information Retrieval

Generative retrieval (GR) reformulates the Information Retrieval (IR) task as the generation of document identifiers (docIDs). Despite its promise, existing GR models exhibit poor generalization to newly added documents, often failing to generate the correct docIDs. While incremental training offers a straightforward remedy, it is computationally expensive, resource-intensive, and prone to catastrophic forgetting, thereby limiting the scalability and practicality of GR. In this paper, we identify the core bottleneck as the decoder's ability to map hidden states to the correct docIDs of newly added documents. Model editing, which enables targeted parameter modifications for docID mapping, represents a promising solution. However, applying model editing to current GR models is not trivial, which is severely hindered by indistinguishable edit vectors across queries, due to the high overlap of shared docIDs in retrieval results. To address this, we propose DOME (docID-oriented model editing), a novel method that effectively and efficiently adapts GR models to unseen documents. DOME comprises three stages: (1) identification of critical layers, (2) optimization of edit vectors, and (3) construction and application of updates. At its core, DOME employs a hybrid-label adaptive training strategy that learns discriminative edit vectors by combining soft labels, which preserve query-specific semantics for distinguishable updates, with hard labels that enforce precise mapping modifications. Experiments on widely used benchmarks, including NQ and MS MARCO, show that our method significantly improves retrieval performance on new documents while maintaining effectiveness on the original collection. Moreover, DOME achieves this with only about 60% of the training time required by incremental training, considerably reducing computational cost and enabling efficient, frequent model updates.

Truth as a Trajectory: What Internal Representations Reveal About Large Language Model Reasoning

Existing explainability methods for Large Language Models (LLMs) typically treat hidden states as static points in activation space, assuming that correct and incorrect inferences can be separated using representations from an individual layer. However, these activations are saturated with polysemantic features, leading to linear probes learning surface-level lexical patterns rather than underlying reasoning structures. We introduce Truth as a Trajectory (TaT), which models the transformer inference as an unfolded trajectory of iterative refinements, shifting analysis from static activations to layer-wise geometric displacement. By analyzing displacement of representations across layers, TaT uncovers geometric invariants that distinguish valid reasoning from spurious behavior. We evaluate TaT across dense and Mixture-of-Experts (MoE) architectures on benchmarks spanning commonsense reasoning, question answering, and toxicity detection. Without access to the activations themselves and using only changes in activations across layers, we show that TaT effectively mitigates reliance on static lexical confounds, outperforming conventional probing, and establishes trajectory analysis as a complementary perspective on LLM explainability.

CM2: Reinforcement Learning with Checklist Rewards for Multi-Turn and Multi-Step Agentic Tool Use

AI agents are increasingly used to solve real-world tasks by reasoning over multi-turn user interactions and invoking external tools. However, applying reinforcement learning to such settings remains difficult: realistic objectives often lack verifiable rewards and instead emphasize open-ended behaviors; moreover, RL for multi-turn, multi-step agentic tool use is still underexplored; and building and maintaining executable tool environments is costly, limiting scale and coverage. We propose CM2, an RL framework that replaces verifiable outcome rewards with checklist rewards. CM2 decomposes each turn's intended behavior into fine-grained binary criteria with explicit evidence grounding and structured metadata, turning open-ended judging into more stable classification-style decisions. To balance stability and informativeness, our method adopts a strategy of sparse reward assignment but dense evaluation criteria. Training is performed in a scalable LLM-simulated tool environment, avoiding heavy engineering for large tool sets. Experiments show that CM2 consistently improves over supervised fine-tuning. Starting from an 8B Base model and training on an 8k-example RL dataset, CM2 improves over the SFT counterpart by 8 points on tau^-Bench, by 10 points on BFCL-V4, and by 12 points on ToolSandbox. The results match or even outperform similarly sized open-source baselines, including the judging model. CM2 thus provides a scalable recipe for optimizing multi-turn, multi-step tool-using agents without relying on verifiable rewards. Code provided by the open-source community: https://github.com/namezhenzhang/CM2-RLCR-Tool-Agent.

CA-YOLO: Cross Attention Empowered YOLO for Biomimetic Localization

In modern complex environments, achieving accurate and efficient target localization is essential in numerous fields. However, existing systems often face limitations in both accuracy and the ability to recognize small targets. In this study, we propose a bionic stabilized localization system based on CA-YOLO, designed to enhance both target localization accuracy and small target recognition capabilities. Acting as the "brain" of the system, the target detection algorithm emulates the visual focusing mechanism of animals by integrating bionic modules into the YOLO backbone network. These modules include the introduction of a small target detection head and the development of a Characteristic Fusion Attention Mechanism (CFAM). Furthermore, drawing inspiration from the human Vestibulo-Ocular Reflex (VOR), a bionic pan-tilt tracking control strategy is developed, which incorporates central positioning, stability optimization, adaptive control coefficient adjustment, and an intelligent recapture function. The experimental results show that CA-YOLO outperforms the original model on standard datasets (COCO and VisDrone), with average accuracy metrics improved by 3.94%and 4.90%, respectively.Further time-sensitive target localization experiments validate the effectiveness and practicality of this bionic stabilized localization system.

Group-Evolving Agents: Open-Ended Self-Improvement via Experience Sharing

Open-ended self-improving agents can autonomously modify their own structural designs to advance their capabilities and overcome the limits of pre-defined architectures, thus reducing reliance on human intervention. We introduce Group-Evolving Agents (GEA), a new paradigm for open-ended self-improvements, which treats a group of agents as the fundamental evolutionary unit, enabling explicit experience sharing and reuse within the group throughout evolution. Unlike existing open-ended self-evolving paradigms that adopt tree-structured evolution, GEA overcomes the limitation of inefficient utilization of exploratory diversity caused by isolated evolutionary branches. We evaluate GEA on challenging coding benchmarks, where it significantly outperforms state-of-the-art self-evolving methods (71.0% vs. 56.7% on SWE-bench Verified, 88.3% vs. 68.3% on Polyglot) and matches or exceeds top human-designed agent frameworks (71.8% and 52.0% on two benchmarks, respectively). Analysis reveals that GEA more effectively converts early-stage exploratory diversity into sustained, long-term progress, achieving stronger performance under the same number of evolved agents. Furthermore, GEA exhibits consistent transferability across different coding models and greater robustness, fixing framework-level bugs in 1.4 iterations on average, versus 5 for self-evolving methods.

ERNIE 5.0 Technical Report

In this report, we introduce ERNIE 5.0, a natively autoregressive foundation model desinged for unified multimodal understanding and generation across text, image, video, and audio. All modalities are trained from scratch under a unified next-group-of-tokens prediction objective, based on an ultra-sparse mixture-of-experts (MoE) architecture with modality-agnostic expert routing. To address practical challenges in large-scale deployment under diverse resource constraints, ERNIE 5.0 adopts a novel elastic training paradigm. Within a single pre-training run, the model learns a family of sub-models with varying depths, expert capacities, and routing sparsity, enabling flexible trade-offs among performance, model size, and inference latency in memory- or time-constrained scenarios. Moreover, we systematically address the challenges of scaling reinforcement learning to unified foundation models, thereby guaranteeing efficient and stable post-training under ultra-sparse MoE architectures and diverse multimodal settings. Extensive experiments demonstrate that ERNIE 5.0 achieves strong and balanced performance across multiple modalities. To the best of our knowledge, among publicly disclosed models, ERNIE 5.0 represents the first production-scale realization of a trillion-parameter unified autoregressive model that supports both multimodal understanding and generation. To facilitate further research, we present detailed visualizations of modality-agnostic expert routing in the unified model, alongside comprehensive empirical analysis of elastic training, aiming to offer profound insights to the community.

Questioning the Coverage-Length Metric in Conformal Prediction: When Shorter Intervals Are Not Better

Conformal prediction (CP) has become a cornerstone of distribution-free uncertainty quantification, conventionally evaluated by its coverage and interval length. This work critically examines the sufficiency of these standard metrics. We demonstrate that the interval length might be deceptively improved through a counter-intuitive approach termed Prejudicial Trick (PT), while the coverage remains valid. Specifically, for any given test sample, PT probabilistically returns an interval, which is either null or constructed using an adjusted confidence level, thereby preserving marginal coverage. While PT potentially yields a deceptively lower interval length, it introduces practical vulnerabilities: the same input can yield completely different prediction intervals across repeated runs of the algorithm. We formally derive the conditions under which PT achieves these misleading improvements and provides extensive empirical evidence across various regression and classification tasks. Furthermore, we introduce a new metric interval stability which helps detect whether a new CP method implicitly improves the length based on such PT-like techniques.

Efficient Multimodal Planning Agent for Visual Question-Answering

Visual Question-Answering (VQA) is a challenging multimodal task that requires integrating visual and textual information to generate accurate responses. While multimodal Retrieval-Augmented Generation (mRAG) has shown promise in enhancing VQA systems by providing more evidence on both image and text sides, the default procedure that addresses VQA queries, especially the knowledge-intensive ones, often relies on multi-stage pipelines of mRAG with inherent dependencies. To mitigate the inefficiency limitations while maintaining VQA task performance, this paper proposes a method that trains a multimodal planning agent, dynamically decomposing the mRAG pipeline to solve the VQA task. Our method optimizes the trade-off between efficiency and effectiveness by training the agent to intelligently determine the necessity of each mRAG step. In our experiments, the agent can help reduce redundant computations, cutting search time by over 60\% compared to existing methods and decreasing costly tool calls. Meanwhile, experiments demonstrate that our method outperforms all baselines, including a Deep Research agent and a carefully designed prompt-based method, on average over six various datasets. Code will be released.

Self-Supervised Path Planning in Unstructured Environments via Global-Guided Differentiable Hard Constraint Projection

Deploying deep learning agents for autonomous navigation in unstructured environments faces critical challenges regarding safety, data scarcity, and limited computational resources. Traditional solvers often suffer from high latency, while emerging learning-based approaches struggle to ensure deterministic feasibility. To bridge the gap from embodied to embedded intelligence, we propose a self-supervised framework incorporating a differentiable hard constraint projection layer for runtime assurance. To mitigate data scarcity, we construct a Global-Guided Artificial Potential Field (G-APF), which provides dense supervision signals without manual labeling. To enforce actuator limitations and geometric constraints efficiently, we employ an adaptive neural projection layer, which iteratively rectifies the coarse network output onto the feasible manifold. Extensive benchmarks on a test set of 20,000 scenarios demonstrate an 88.75\% success rate, substantiating the enhanced operational safety. Closed-loop experiments in CARLA further validate the physical realizability of the planned paths under dynamic constraints. Furthermore, deployment verification on an NVIDIA Jetson Orin NX confirms an inference latency of 94 ms, showing real-time feasibility on resource-constrained embedded hardware. This framework offers a generalized paradigm for embedding physical laws into neural architectures, providing a viable direction for solving constrained optimization in mechatronics. Source code is available at: https://github.com/wzq-13/SSHC.git.

The Geometric Mechanics of Contrastive Representation Learning: Alignment Potentials, Entropic Dispersion, and Cross-Modal Divergence

While InfoNCE powers modern contrastive learning, its geometric mechanisms remain under-characterized beyond the canonical alignment--uniformity decomposition. We present a measure-theoretic framework that models learning as the evolution of representation measures on a fixed embedding manifold. By establishing value and gradient consistency in the large-batch limit, we bridge the stochastic objective to explicit deterministic energy landscapes, uncovering a fundamental geometric bifurcation between the unimodal and multimodal regimes. In the unimodal setting, the intrinsic landscape is strictly convex with a unique Gibbs equilibrium; here, entropy acts merely as a tie-breaker, clarifying "uniformity" as a constrained expansion within the alignment basin. In contrast, the symmetric multimodal objective contains a persistent negative symmetric divergence term that remains even after kernel sharpening. We show that this term induces barrier-driven co-adaptation, enforcing a population-level modality gap as a structural geometric necessity rather than an initialization artifact. Our results shift the analytical lens from pointwise discrimination to population geometry, offering a principled basis for diagnosing and controlling distributional misalignment.