The article introduces the ICML 2026 Theory of Agent (ToA), analyzes four common failure modes of modern agents, explains the internal‑vs‑external tool trade‑off through a knowledge‑boundary framework, and outlines how effort‑conservation and the β parameter guide self‑evolving agent design and future research.
The paper introduces Language Model Networks (LMNet), a framework that lets pretrained large language models act as reusable compute nodes communicating via dense, trainable vectors, showing measurable performance gains on general and supervised adaptation tasks with minimal extra training cost.
The paper presents MP‑MoE, a new Mixture‑of‑Experts architecture that replaces top‑k routing with Mahalanobis‑based ensemble pruning, explicitly encouraging expert diversity via a co‑occurrence matrix, and uses an efficient greedy algorithm with incremental Cholesky updates, achieving higher performance with minimal training overhead and no inference cost.
The paper introduces PACE, a Parameter‑Change based Unsupervised Environment Design method that evaluates training levels by the magnitude of induced policy‑parameter updates, offering a low‑variance, computationally cheap signal that consistently outperforms prior UED approaches on MiniGrid and Craftax benchmarks.
The paper introduces PACE, a Parameter Change Environment Design method that evaluates training levels by measuring induced policy parameter updates, offering a low‑variance learning‑progress signal that outperforms prior UED approaches on MiniGrid and Craftax benchmarks, achieving higher success rates and more stable generalization.
APO (Autonomous Preference Optimization) converts the drift and conflict among multiple teacher multimodal LLMs into dynamic negative constraints while treating consensus as a positive preference, enabling robust concept alignment and superior diagnostic accuracy on the CXR‑MAX benchmark, as demonstrated by extensive ICML‑2026 experiments.
The paper introduces APO, a novel autonomous preference optimization framework that converts concept drift among multiple teacher multimodal LLMs into dynamic negative constraints and treats consensus as a positive preference, achieving robust concept alignment and surpassing strong teachers on a high‑risk medical X‑ray benchmark.
Reviewers discovered hidden text in ICML 2026 PDFs that injects specific phrases into large‑language‑model generated reviews, turning an attack technique into a defense mechanism and prompting new safeguards such as watermarking and OCR‑based checks.
ICML 2026 opens paper submissions with strict AI usage rules—LLMs cannot be listed as authors, prompt injection is banned, and AI reviewing is expanded—while outlining submission formats, important dates, reciprocal review limits, and ethical guidelines for authors.
