Latent test-time scaling improves reasoning by refining hidden states during inference, but existing methods typically apply a single scalar reward to all editable latent tokens. For multimodal large language models, this global update ignores that generated tokens play different roles: some are sensitive to visual evidence, while others correspond to uncertain reasoning decisions. We present Token-Disentangled Latent Test-Time Scaling, an inference-time framework that makes latent refinement token-role-aware. Starting from an initial generated trajectory, we optimize a short hidden-state prefix while routing perception-side visual feedback to image-sensitive tokens and reasoning feedback to high-entropy tokens. Tokens selected by neither route are constrained by an anchor regularizer. Across six multimodal reasoning benchmarks and two frozen MLLM backbones, our method consistently improves over the initial chain-of-thought rollout: macro accuracy increases from 66.38 to 68.95 on Qwen2.5-VL-7B and from 67.93 to 69.44 on InternVL3.5-8B. These results show that token-disentangled latent refinement is an effective strategy for multimodal test-time scaling and remains competitive with output-space test-time scaling baselines.

Coding agents working on sequential maintenance tasks re-read the same files across related tasks, wasting more than a third of their input tokens on redundant context. The natural fix, caching file summaries, fails to reliably reduce cost: tool-use traces are consistent with agents over-relying on generic descriptions, skipping re-reads, and offsetting token savings with less effective downstream decisions, leaving net cost roughly unchanged. We introduce Prism, a content-hash-keyed file memory that augments summaries with bounded facets: identifier-grounded facts distilled from the agent’s own prior reasoning. On 12 SWE-Chain dependency-upgrade chains, Prism reduces cost by 24.9% on GPT-5 and 9.9% on GPT-5.4-mini, while maintaining comparable aggregate Build+Fix F1. Against no memory, a summary-only ablation yields only a nominal 2.4% cost decrease that falls within run-to-run variability, while a static Aider-style repo-map baseline costs substantially more at 13% over no memory. Two controlled ablations isolate the causal contributions: wiping the persistent store before every task raises cost by 37% over no memory, and replacing facets with a 3× larger unstructured summary raises it by 41%. Code: https://github.com/WhiteFlamesCN/Prism-Opencode.

Post-training large language models often require optimizing multiple objectives simultaneously, motivating multi-reward reinforcement learning (RL) methods that model different objectives as separate reward signals. Existing methods like Group reward-Decoupled Policy Optimization (GDPO) compute group-normalized advantages separately for each reward dimension and then aggregate them for policy updates. However, this aggregation process can cause cross-reward advantage conflict: the same rollout may have positive advantages on some reward dimensions but negative advantages on others, causing positive and negative signals to cancel each other during aggregation. To address this issue, we propose GD²PO, short for Group-Dynamic Reward-Decoupled Policy Optimization, a conflict-aware multi-reward policy optimization method. Before advantage aggregation, GD²PO uses conflict-aware criteria to identify and filter rollouts with severe reward-wise disagreement, reducing the influence of unreliable update signals. It further introduces query-level reweighting to dynamically adjust the update strength of each query according to its overall reward consensus. Experiments on tool calling and helpfulness-safety alignment show that GD²PO outperforms existing methods. Our code is available at https://github.com/Qwen-Applications/GD2PO.

Large language models (LLMs) have shown promise in complex reasoning and tool-based decision making, motivating their application to real-world supply chain management. However, supply chain workflows require reliable long-horizon, multi-step orchestration grounded in domain-specific procedures, which remains challenging for current models. To systematically evaluate LLM performance in this setting, we introduce SupChain-Bench, a unified real-world benchmark that assesses both supply chain domain knowledge and long-horizon tool-based orchestration grounded in standard operating procedures (SOPs). Our experiments reveal substantial gaps in execution reliability across models. We further propose SupChain-ReAct, an SOP-free framework that autonomously synthesizes executable procedures for tool use, achieving the strongest and most consistent tool-calling performance. Our work establishes a principled benchmark for studying reliable long-horizon orchestration in real-world operational settings and highlights significant room for improvement in LLM-based supply chain agents.

Large Language Model (LLM) agents increasingly rely on domain-specific skills, yet manually authoring such skills does not scale, and skills generated purely from parametric knowledge often miss critical operational pitfalls. We introduce Trace2Skill, a framework that consolidates broad execution trajectories in parallel into a unified skill directory through inductive reasoning over agent experience. Trace2Skill supports both deepening existing human-written skills and creating useful skills from weak LLM-generated drafts. Experiments demonstrate the effectiveness of Trace2Skill across diverse domains, including office workflows, math reasoning, and vision QA. Importantly, the evolved skills are not merely memorized artifacts of the trajectories used to create them: they often transfer across model scales, across model families, and to out-of-distribution settings. For example, skills evolved from Qwen3.5-35B trajectories improve a Qwen3.5-122B agent by up to 57.65 percentage points on WikiTableQuestions. Further analyses show that Trace2Skill outperforms sequential skill editing and ReasoningBank-style retrieval memories, compresses recurring failures and workarounds into standard operating procedures (SoPs), and yields portable skills that can be reused without parameter updates or test-time retrieval.

Tabular data are crucial in many fields and their understanding by large language models (LLMs) under high parameter efficiency paradigm is important. However, directly applying parameter-efficient fine-tuning (PEFT) techniques to tabular tasks presents significant challenges, particularly in terms of better table serialization and the representation of two-dimensional structured information within a one-dimensional sequence. To address this, we propose TableLoRA, a module designed to improve LLMs’ understanding of table structure during PEFT. It incorporates special tokens for serializing tables with special token encoder and uses 2D LoRA to encode low-rank information on cell positions. Experiments on four tabular-related datasets demonstrate that TableLoRA consistently outperforms vanilla LoRA and surpasses various table encoding methods tested in control experiments. These findings reveal that TableLoRA, as a table-specific LoRA, enhances the ability of LLMs to process tabular data effectively, especially in low-parameter settings, demonstrating its potential as a robust solution for handling table-related tasks.

Tabular data analysis is crucial in many scenarios, yet efficiently identifying relevant queries and results for new tables remains challenging due to data complexity, diverse analytical operations, and high-quality analysis requirements. To address these challenges, we aim to recommend query–code–result triplets tailored for new tables in tabular data analysis workflows. In this paper, we present TablePilot, a pioneering tabular data analysis framework leveraging large language models to autonomously generate comprehensive and superior analytical results without relying on user profiles or prior interactions. Additionally, we propose Rec-Align, a novel method to further improve recommendation quality and better align with human preferences. Experiments on DART, a dataset specifically designed for comprehensive tabular data analysis recommendation, demonstrate the effectiveness of our framework. Based on GPT-4o, the tuned TablePilot achieves 77.0% top-5 recommendation recall. Human evaluations further highlight its effectiveness in optimizing tabular data analysis workflows.