LLM Agents

MASCOT is a multi-agent socio-collaborative companion framework that coordinates specialized agents around social context and user goals to enable trustworthy, everyday companion experiences.

Jul 1, 2026

Hierarchical agentic framework for multi-page visual document understanding. Decomposes reasoning into global, page, and element levels to handle slide decks, financial reports, and infographics. Accepted at ACL 2026 main conference.

Mar 15, 2026

AgentArk distills the collaborative behavior of multi-agent systems into a single LLM agent, decomposing trajectories into role-conditioned skills and recovering most of the ensemble's performance at a fraction of the cost.

Feb 4, 2026

Distilling multi-agent intelligence into a single LLM agent. Decomposes multi-agent trajectories into role-conditioned skills and trains a single agent to reproduce the collaborative behavior of the original ensemble. Under review at NeurIPS 2026.

Feb 4, 2026

Multi-page visual documents such as manuals, brochures, presentations, and posters convey key information through layout, colors, icons, and cross-slide references. While large language models (LLMs) offer opportunities in document understanding, current systems struggle with complex, multi-page visual documents, particularly in fine-grained reasoning over elements and pages. We introduce SlideAgent, a versatile agentic framework for understanding multi-modal, multi-page, and multi-layout documents, especially slide decks. SlideAgent employs specialized agents and decomposes reasoning into three specialized levels--global, page, and element--to construct a structured, query-agnostic representation that captures both overarching themes and detailed visual or textual cues. During inference, SlideAgent selectively activates specialized agents for multi-level reasoning and integrates their outputs into coherent, context-aware answers. Extensive experiments show that SlideAgent achieves significant improvement over both proprietary (+7.9 over GPT-4o) and open-source models (+9.8 over InternVL3-8B).

Sep 26, 2025

Peer review is fundamental to the integrity and advancement of scientific publication. Traditional methods of peer review analyses often rely on exploration and statistics of existing peer review data, which do not adequately address the multivariate nature of the process, account for the latent variables, and are further constrained by privacy concerns due to the sensitive nature of the data. We introduce AgentReview, the first large language model (LLM) based peer review simulation framework, which effectively disentangles the impacts of multiple latent factors and addresses the privacy issue. Our study reveals significant insights, including a notable 37.1% variation in paper decisions due to reviewers' biases, supported by sociological theories such as the social influence theory, altruism fatigue, and authority bias. We believe that this study could offer valuable insights to improve the design of peer review mechanisms. Our code is available at https://github.com/Ahren09/AgentReview.

Nov 12, 2024

Peer review is fundamental to the integrity and advancement of scientific publication. Traditional methods of peer review analyses often rely on exploration and statistics of existing peer review data...

Nov 12, 2024

First LLM-based peer review simulation framework that disentangles latent factors driving reviewer decisions. Reveals a 37.1% variation in paper decisions due to reviewer biases. EMNLP 2024 Oral.

Nov 12, 2024

Studies competition dynamics among LLM-based agents in a simulated virtual town with restaurant and customer agents. Reveals emergent behaviors and strategic patterns aligned with market and sociological theories. ICML 2024 Oral.

May 1, 2024

This work studies the competition dynamics among LLM-based agents, revealing emergent behaviors and strategic patterns in multi-agent systems....

Apr 30, 2024