The paper introduces ProAS, a Progress‑Aware Advantage Shaping framework designed to improve critic‑free reinforcement learning for large language model agents. ProAS uses a dynamic Agentic Reference Graph to provide fine‑grained credit assignment through Masking Unreasonable Incentives and Asymmetric Penalty Redistribution, avoiding the uniform advantage broadcast that can reinforce invalid actions. Experiments on Qwen2.5‑1.5B‑Instruct and Qwen2.5‑7B‑Instruct show that ProAS raises baseline success rates to 91.9% and 94.3% on ALFWorld and 66.7% and 70.8% on WebShop, while speeding up training convergence up to 2.44× and remaining robust to imperfect reference graphs.
By Bowen Ding, Mingda Wan, Yuhan Chen, Jiyao Yuan, Qi Zhu, Futing Wang, Heyuan Deng, Fei Mi, Lifeng Shang, Tao Lin
arXiv:2610.16067v1 Announce Type: cross
Abstract: Agent evaluation harnesses are software systems that run test suites, generate inputs, and decide outcomes with test oracles. This paper audits such...
By Junjie Chen
Large language models increasingly rely on external memory to answer queries grounded in long-horizon histories such as multi-session dialogues and lengthy documents. However, existing memory methods...
LLM-agent evaluation reports behavioral scores but leaves the identity of the deployed runtime stack unreported. A tampered deployment that preserves evaluated outputs passes every output-only evaluat...
Tool: LaTeX renderer
Inspired by Paul Ford and Aboard Foundary's new web app Tiny Little Office, which provides a suite of web app content editors (a Word Processor, Outliner, Presentation Bui...
We present CommerceGen-Studio, an end-to-end framework for generating commercial-grade main images for e-commerce Stock Keeping Units (SKUs). Our framework is organized along a three-stage pipeline of...
We introduce a novel method for verifying Lean formalizations that goes beyond LLM-judged faithfulness. Unlike most verification methods, which rely on LLMs or trained models, our method combines LLM-...
3D object composition is pivotal to digital content creation and robot simulation, while also serving as a critical testbed for physical intelligence in foundation models. Previous methods either leve...
Understanding the social intelligence of large language models (LLMs) is increasingly important for applications involving human-AI interaction and multi-agent environments. Social intelligence compri...
The article discusses how AI is enabling large-scale decompilation of games, allowing developers to create mashups and preserve old titles by porting them to modern platforms like WebAssembly. It highlights both the exciting possibilities for game preservation and the frightening implications for intellectual property, as anyone could decompile and repurpose software without permission. The piece underscores the dual nature of this technology—boosting accessibility and preservation while threatening the security of proprietary software.
With the rapid development of vision-language models (VLMs), Web-GUI agents have achieved promising performance in interacting with real-world interfaces through screenshot-based grounding. However, h...
Agents for mixed-degradation image restoration typically learn what a tool does to an image only after running it: failed calls waste computation and harmful outputs go unnoticed until too late. We pr...
Agents driven by large language models are increasingly deployed on long-horizon tasks, where they use tools and modify real workspaces. How well they perform depends as much on the harness around the...
A ReAct-style agent accumulates observations, intermediate results, and failed attempts in a single context. As this context grows, stale and off-path information consumes its limited capacity and obs...
Effective social interaction requires more than fluent responses: agents must reach plans that fit the participants' needs and circumstances. Existing evaluations of social agents mainly rely on LLM j...
Medication recommendation (MR) is critical for clinical decision support. While existing MR methods are mostly designed for single-center learning, where only one independent hospital is involved, mul...
Graphical user interface (GUI) agents now execute workflows across real applications, but when execution drifts, users must determine what changed, what should have remained untouched, and what can st...
Amp demonstrates the expanding capabilities of AI agents by showcasing their use with OpenAI models such as GPT-6 Astra, Ultrafast, GPT-Live-1, and Sign in with ChatGPT. The article highlights how these agents can perform tasks beyond traditional coding, illustrating the broader applicability of GPT-based technologies. It provides examples of new functionalities enabled by these advanced models.
Memory systems for persistent personal agents must move beyond isolated memories, maintaining both personal memory about whom they serve and procedural memory about how to act. Crucially, in harness-b...
The article "A Scaffold at the Gate: Building AI Applications That Can Evolve" reflects on the necessity for data scientists to adopt software engineering principles in the context of AI agents that generate code. It discusses how these principles can help create AI applications capable of evolving over time. The piece was originally published on Towards Data Science.
By Erika Gomes-Gonçalves