❌

Normal view

Received — 27 May 2026 ⏭

Benchmarking the Limits of In-Context Reinforcement Learning for Ad-Hoc Teamwork

✇cs.AI, q-bio.NC updates on arXiv.org
By: Yuheng Jing · Kai Li · Ziwen Zhang · Jiajun Zhang · Zeyao Ma · Jiaxi Yang · Lei Zhang · Zhe Wu · Jinmin He · Junliang Xing · Jian Cheng
26 May 2026 at 12:00
arXiv:2605.24423v1 Announce Type: new Abstract: In-Context Reinforcement Learning (ICRL) has enabled foundation agents to adapt instantaneously to novel tasks, yet its efficacy in Ad-Hoc Teamwork (AHT)-where coordination with unknown partners is required-remains unexplored. To rigorously evaluate this, we introduce a large-scale benchmark ICRL4AHT, built upon a high-throughput JAX implementation of Overcooked-V2. Our benchmark includes a large, diverse teammate suite spanning both RL and heuristic policies, enabling controlled train-test shifts, and provides a reproducible end-to-end pipeline for teammate generation, learning-history collection, dataset construction, and online multi-episode evaluation. We evaluate representative history-conditioned ICRL algorithms, including Algorithm Distillation (AD) and Decision-Pretrained Transformer (DPT), across millions of transitions. Results reveal notable limitations: contrary to their success in single-agent domains, these baselines fail to exhibit robust test-time adaptation in multi-agent settings. Specifically, these methods frequently underperform random baselines across both unseen teammate and unseen layout tracks, with no clear in-context improvement over long horizons. These findings highlight the challenges of strategic inference under partial observability within the OvercookedV2 AHT protocol, establishing our benchmark as a critical testbed for next-generation coordination algorithms.

Test-Time Deep Thinking to Explore Implicit Rules

✇cs.AI, q-bio.NC updates on arXiv.org
By: Wentong Chen · Xin Cong · Zhong Zhang · Yaxi Lu · Siyuan Zhao · Yesai Wu · Qinyu Luo · Haotian Chen · Yankai Lin · Zhiyuan Liu · Maosong Sun
26 May 2026 at 12:00
arXiv:2605.24828v1 Announce Type: new Abstract: With the continuous advancement of Large Language Models (LLMs), intelligent agents are becoming increasingly vital. However, these agents often fail in environments governed by implicit rules--hidden constraints that cannot be observed directly and must be inferred through interaction. This causes agents to fall into repetitive trial-and-error loops, ultimately leading to task failure. To address this challenge, we propose Test-Time Exploration (TTExplore), a framework where a thinker component analyzes interaction history to infer these implicit rules and guide an actor. Effective exploration in this setting critically depends on the reasoning ability of the thinker. However, evaluating deep reasoning trajectories is inherently unstable and difficult, which poses a major obstacle to effective training. To overcome this issue, we introduce a novel and stable reinforcement learning pipeline. The core idea is to use accurate task-level scores as indirect rewards to bypass the difficulty of evaluating intermediate reasoning, and to retain only a single thinking node per trajectory to alleviate reward sparsity. Using this pipeline, we train a specialized 7B model, Exp-Thinker. Experiments on five text-based embodied tasks show that TTExplore equipped with Exp-Thinker improves baseline agent performance by an average of $14$-$19$ points, demonstrating the effectiveness of explicitly reasoning about implicit rules.

FrontierOR: Benchmarking LLMs' Capacity for Efficient Algorithm Design in Large-Scale Optimization

✇cs.AI, q-bio.NC updates on arXiv.org
By: Minwei Kong · Chonghe Jiang · Ao Qu · Wenbin Ouyang · Zhaoming Zeng · Xiaotong Guo · Zhekai Li · Junyi Li · Yi Fan · Xinshou Zheng · Xi Jing · Yikai Zhang · Zhiwei Liang · Seonghoo Kim · Runqing Yang · Zijian Zhou · Sirui Li · Han Zheng · Wangyang Ying · Ou Zheng · Chonghuan Wang · Jinglong Zhao · Hanzhang Qin · Cathy Wu · Paul Pu Liang · Jinhua Zhao · Hai Wang
26 May 2026 at 12:00
arXiv:2605.25246v2 Announce Type: new Abstract: Large language models (LLMs) are increasingly used for optimization modeling and solver-code generation, yet practical operations research and optimization problems often require a harder capability: designing scalable algorithms that exploit problem structure and outperform direct formulation-and-solve baselines. Existing benchmarks are limited to small or simplified examples far below real-world scale and complexity. We introduce FrontierOR, among the first benchmarks to systematically evaluate LLM-based efficient algorithm design for realistic large-scale optimization problems. FrontierOR includes 180 tasks derived from methodologically diverse papers published in top-tier operations research venues, each with standardized instances and a hidden, expert-verified evaluation suite. We evaluate seven LLMs spanning frontier, cost-effective, and open-source models both in one-shot and test-time evolution settings. The results reveal that frontier models still struggle to move from executable formulations to efficient optimization algorithms: the strongest one-shot model outperforms Gurobi in only 31% of cases in both solution quality and computational efficiency, and even strong coding agents with test-time evolution achieve only 50% on selected hard tasks. FrontierOR establishes a practical evaluation platform for LLM-based optimization algorithm design, which enables future LLMs and agents to be systematically tested on whether they can move beyond correct formulation toward a feasible, high-quality, and efficient algorithm.

DRScaffold: Boosting Dense-Scene Reasoning in Lightweight Vision Language Models

✇cs.AI, q-bio.NC updates on arXiv.org
By: Xinrui Shi · Kai Liu · Ziqing Zhang · Jianze Li · Anqi Li · Yulun Zhang
26 May 2026 at 12:00
arXiv:2605.26038v1 Announce Type: cross Abstract: Lightweight vision-language models perform competitively on standard benchmarks yet fail systematically in dense-scene reasoning, where multiple objects, attributes, and relations must be jointly grounded and resolved through multi-step inference. Such capability is critical for real-world applications where models must reliably interpret cluttered environments. Yet existing training signals provide no explicit grounding between reasoning steps and the underlying visual entities and relations, leaving lightweight models free to generate fluent but visually unanchored reasoning chains. To address this gap, we first introduce DRBench, a benchmark of 14,573 questions across 2,943 images, organized into five task categories spanning three progressive reasoning layers. Building on DRBench, we propose DRScaffold, a supervised fine-tuning framework that decomposes the supervision target into four causally ordered stages, enforcing grounded reasoning without architectural modification. Experiments on three lightweight VLMs demonstrate substantial gains on DRBench while preserving or improving performance on general-purpose benchmarks. Notably, Qwen2.5-VL-3B trained with DRScaffold surpasses the frozen Qwen2.5-VL-32B on DRBench, demonstrating that structured supervision can substitute for a significant portion of model scale in dense-scene reasoning. Our code and models are available at https://github.com/irene-shi/DRScaffold .

Characterization of dysbiosis patterns in gut microbiota of digestive system cancers: an umbrella review

✇Omics in Gastric
By: Jing Wang · Yingkai Liu · Yan Yu · Xiaoli Chen · Genghang Chen · Yang Chen · Xueyin Chen · Shaonan Liu
14 May 2026 at 18:00

Front Microbiol. 2026 Apr 28;17:1782471. doi: 10.3389/fmicb.2026.1782471. eCollection 2026.

ABSTRACT

Digestive system cancers (DSCs) represent a substantial global health burden. In recent years, the role of gut microbiota in the DSCs has garnered considerable attention, but its change pattern during tumor progression and the specific mechanisms are still not fully understood. We conducted a comprehensive systematic review to characterize patterns of gut microbiota dysbiosis across different DSC types and assess their clinical significance. We systematically searched four English and three Chinese databases up to January 2025 to identify systematic reviews focused on the dynamic characteristics of the gut microbiota during gastrointestinal tumorigenesis. Microbiota biodiversity and taxonomic composition were extracted to identify specific signatures associated with DSCs. The ROBIS tool was used to evaluate the methodological quality of the included studies. Ultimately, 59 studies involving six distinct DSC types were included. Data synthesis and comparison revealed distinct microbiota profiles across DSCs. At the phylum level, Bacillota was decreased in esophageal cancer (EC) and pancreatic ductal adenocarcinoma (PDAC), Pseudomonadota was augmented in EC but exhibited divergent trajectories in colorectal cancer (CRC) and PDAC. Genus-level analyses revealed Veillonella enrichment in EC and PDAC, and Fusobacterium outgrowth in EC, gastric cancer (GC) and CRC. Parvimonas and Streptococcus showed a concordant ascending trend in GC and CRC. Prevotella was overrepresented in EC and GC. This synthesis delineates a qualitative landscape of gut microbiota imbalances associated with various DSCs, highlighting the potential for these microbial shifts to serve as markers for early detection and targeted therapy. Multiomics integration and prospective cohort studies should be prioritized to accelerate clinical translation.

PMID:42131199 | PMC:PMC13161176 | DOI:10.3389/fmicb.2026.1782471

There are no more articles
❌