Select-then-Solve: Paradigm Routing as Inference-Time Optimization for LLM Agents

When an LLM-based agent improves on a task, is the gain from the model itself or from the reasoning paradigm wrapped around it? We study this question by comparing six inference-time paradigms, namely Direct, CoT, ReAct, Plan-Execute, Reflection, and ReCode, across four frontier LLMs and ten benchmarks, yielding roughly 18,000 runs. We find that reasoning structure helps dramatically on some tasks but hurts on others: ReAct improves over Direct by 44pp on GAIA, while CoT degrades performance by 15pp on HumanEval. No single paradigm dominates, and oracle per-task selection beats the best fixed paradigm by 17.1pp on average. Motivated by this complementarity, we propose a select-then-solve approach: before answering each task, a lightweight embedding-based router selects the most suitable paradigm. Across four models, the router improves average accuracy from 47.6% to 53.1%, outperforming the best fixed paradigm at 50.3% by 2.8pp and recovering up to 37% of the oracle gap. In contrast, zero-shot self-routing only works for GPT-5 at 67.1% and fails for weaker models, all trailing the learned router. Our results argue that reasoning paradigm selection should be a per-task decision made by a learned router, not a fixed architectural choice.

CoEnv: Driving Embodied Multi-Agent Collaboration via Compositional Environment

Multi-agent embodied systems hold promise for complex collaborative manipulation, yet face critical challenges in spatial coordination, temporal reasoning, and shared workspace awareness. Inspired by human collaboration where cognitive planning occurs separately from physical execution, we introduce the concept of compositional environment -- a synergistic integration of real-world and simulation components that enables multiple robotic agents to perceive intentions and operate within a unified decision-making space. Building on this concept, we present CoEnv, a framework that leverages simulation for safe strategy exploration while ensuring reliable real-world deployment. CoEnv operates through three stages: real-to-sim scene reconstruction that digitizes physical workspaces, VLM-driven action synthesis supporting both real-time planning with high-level interfaces and iterative planning with code-based trajectory generation, and validated sim-to-real transfer with collision detection for safe deployment. Extensive experiments on challenging multi-arm manipulation benchmarks demonstrate CoEnv's effectiveness in achieving high task success rates and execution efficiency, establishing a new paradigm for multi-agent embodied AI.

Can RL Improve Generalization of LLM Agents? An Empirical Study

Reinforcement fine-tuning (RFT) has shown promise for training LLM agents to perform multi-turn decision-making based on environment feedback. However, most existing evaluations remain largely in-domain: training and testing are conducted in the same environment or even on the same tasks. In real-world deployment, agents may operate in unseen environments with different background knowledge, observation spaces, and action interfaces. To characterize the generalization profile of RFT under such shifts, we conduct a systematic study along three axes: (1) within-environment generalization across task difficulty, (2) cross-environment transfer to unseen environments, and (3) sequential multi-environment training to quantify transfer and forgetting. Our results show that RFT generalizes well across task difficulty within an environment, but exhibits weaker transfer to unseen environments, which correlates with shifts in both semantic priors and observation/action interfaces. In contrast, sequential training yields promising downstream gains with minimal upstream forgetting, and mixture training across environments improves the overall balance. We further provide detailed analyses and deeper insights, and hope our work helps the community develop and deploy generalizable LLM agents.

Reading $ eq$ Seeing: Diagnosing and Closing the Typography Gap in Vision-Language Models

Vision-Language Models achieve near-perfect accuracy at reading text in images, yet prove largely typography-blind: capable of recognizing what text says, but not how it looks. We systematically investigate this gap by evaluating font family, size, style, and color recognition across 26 fonts, four scripts, and three difficulty levels. Our evaluation of 15 state-of-the-art VLMs reveals a striking perception hierarchy: color recognition is near-perfect, yet font style detection remains universally poor. We further find that model scale fails to predict performance and that accuracy is uniform across difficulty levels, together pointing to a training-data omission rather than a capacity ceiling. LoRA fine-tuning on a small set of synthetic samples substantially improves an open-source model, narrowing the gap to the best closed-source system and surpassing it on font size recognition. Font style alone remains resistant to fine-tuning, suggesting that relational visual reasoning may require architectural innovation beyond current patch-based encoders. We release our evaluation framework, data, and fine-tuning recipe to support progress in closing the typographic gap in vision-language understanding.

SciAgentGym: Benchmarking Multi-Step Scientific Tool-use in LLM Agents

Scientific reasoning inherently demands integrating sophisticated toolkits to navigate domain-specific knowledge. Yet, current benchmarks largely overlook agents' ability to orchestrate tools for such rigorous workflows. To bridge this gap, we introduce SciAgentGym, a scalable interactive environment featuring 1,780 domain-specific tools across four natural science disciplines, supported by a robust execution infrastructure. Complementing this, we present SciAgentBench, a tiered evaluation suite designed to stress-test agentic capabilities from elementary actions to long-horizon workflows. Our evaluation identifies a critical bottleneck: state-of-the-art models struggle with complex scientific tool-use. Even for a leading model like GPT-5, success rates drop sharply from 60.6% to 30.9% as interaction horizons extend, primarily due to failures in multi-step workflow execution. To address this, we propose SciForge, a data synthesis method that models the tool action space as a dependency graph to generate logic-aware training trajectories. By fine-tuning on these trajectories, our SciAgent-8B outperforms the significantly larger Qwen3-VL-235B-Instruct while exhibiting positive cross-domain transfer of scientific tool-use capabilities. These results underscore the promising potential of next-generation autonomous scientific agents.

Visual-O1: Understanding Ambiguous Instructions via Multi-modal Multi-turn Chain-of-thoughts Reasoning

As large-scale models evolve, language instructions are increasingly utilized in multi-modal tasks. Due to human language habits, these instructions often contain ambiguities in real-world scenarios, necessitating the integration of visual context or common sense for accurate interpretation. However, even highly intelligent large models exhibit significant performance limitations on ambiguous instructions, where weak reasoning abilities of disambiguation can lead to catastrophic errors. To address this issue, this paper proposes Visual-O1, a multi-modal multi-turn chain-of-thought reasoning framework. It simulates human multi-modal multi-turn reasoning, providing instantial experience for highly intelligent models or empirical experience for generally intelligent models to understand ambiguous instructions. Unlike traditional methods that require models to possess high intelligence to understand long texts or perform lengthy complex reasoning, our framework does not significantly increase computational overhead and is more general and effective, even for generally intelligent models. Experiments show that our method not only significantly enhances the performance of models of different intelligence levels on ambiguous instructions but also improves their performance on general datasets. Our work highlights the potential of artificial intelligence to work like humans in real-world scenarios with uncertainty and ambiguity. We will release our data and code.