Sichuan University researchers propose RTWI, a plug-and-play test-time scaling framework that models reasoning reliability to filter noisy thinking in multimodal chain-of-thought, achieving higher accuracy and efficiency on benchmarks like Qwen3-VL and DeepEyes.
Tsinghua researchers propose Effective Degree (ED), a differentiable measure of neural network simplicity based on polynomial representations along data interpolation paths, which correlates strongly with generalization gap and can be used as a regularizer to improve performance across vision, language, and RL tasks.
InstEmb introduces learnable look‑ahead tokens and dual‑anchor alignment pooling to inject output‑aware semantics into instruction‑following embeddings without extra decoding, achieving strong gains on ICML‑2026 benchmarks such as FollowIR and InfoSearch while maintaining competitive performance on general sentence‑embedding tasks.
At ICML 2026, ten closed‑door sessions with leading researchers uncovered emerging signals—from next‑generation diffusion language models and data‑centric AI to AI‑driven finance, autonomous agents, AI as an operating system, and AI for science—highlighting the directions that will shape AI research and deployment over the next few years.
A paper accepted to ICML 2026 introduces Verbalized Sampling, a prompt‑only technique that dramatically improves large‑language‑model output diversity by addressing mode collapse through typicality bias, achieving 1.6–2.1× more varied generations without sacrificing accuracy, while igniting polarized discussion on Reddit.
This article reviews three ICML 2026 papers that introduce a user‑centric benchmark for mobile GUI agents, a path‑likelihood optimizer for diffusion language models, and a graph‑structured framework that prunes training datasets without sacrificing accuracy.
ICML 2026 in Seoul broke submission records, sparked controversy over LLM‑generated reviews, honored breakthrough papers on diffusion models, reinforcement learning and AI alignment, and culminated in a reflective question about the true purpose and beneficiaries of machine learning.
ICML 2026 announced ten award‑winning papers, highlighting novel insights such as the flexibility trap in diffusion language models, high‑accuracy sampling for diffusion, risks of AI alignment tools, a random‑matrix view of diffusion consistency, grokking in ridge regression, and an asynchronous deep‑RL framework, each accompanied by concise abstracts and links.
The paper reveals that failures in LLM‑driven multi‑agent systems often stem from the Orchestrator’s loss of control, introduces an entropy‑dynamics framework to measure scheduling entropy, and proposes Inverse Workflow Generation for detailed process evaluation, shifting focus from agent strength to orchestration stability.
The MetaphorVU project introduces the first systematic benchmark for metaphor video understanding, builds a taxonomy of eight metaphor types from billions of real short videos, evaluates 11 multimodal LLMs revealing a 20‑point gap to human performance, and proposes MetaphorBoost—a knowledge‑graph‑enhanced inference framework that consistently improves metaphor comprehension across models.
The ICML 2026 paper reframes in‑context learning as approximate Bayesian inference, introduces explicit prior datasets as a context prefix for Transformers, and demonstrates through synthetic and real‑world experiments that this multi‑task approach closely matches Bayesian oracles while offering fast, controllable inference.
The paper identifies routing drift and expert drift as the two main causes of forgetting in multimodal continual instruction tuning (MCIT) and proposes SAME, which combines spectral‑aware routing, curvature‑aware scaling, and adaptive expert activation to keep MoE models stable, efficient, and less forgetful across long task sequences.
The ICML 2026 spotlight paper argues that universal aesthetic alignment in image‑generation models narrows artistic expression, presents six interrelated concerns, and demonstrates through extensive prompts and benchmark tests that reward models and aligned generators stubbornly favor homogenized, overly positive imagery while failing to honor anti‑aesthetic or negative‑emotion requests.
Holi‑Spatial introduces a fully automatic pipeline that transforms raw video streams into high‑quality 3D geometry, depth, masks, 3D boxes, instance descriptions, grounding and spatial QA, producing the 4‑million‑item Holi‑Spatial‑4M dataset and substantially improving VLM spatial reasoning performance.
The Kuaishou technology team reports that several of its papers were accepted at the prestigious ICML 2026 conference—including a spotlight paper on metaphor video understanding, works on causal discovery for irregular time series, image super‑resolution, large‑scale notification dispatch, full‑order ranking, phase‑aware MoE for RL, end‑to‑end e‑commerce search, spatial‑reasoning rewards, a unified SWE benchmark, video temporal grounding, and interpretable transformers—while also inviting attendees to visit their booth B101 in Seoul.
A joint effort by Baidu Baige and Fudan University introduces the LU‑KV framework, which treats KV‑cache budget allocation as a global combinatorial optimization problem, achieving only 0.52% relative performance loss at 80% compression and establishing a new efficiency‑accuracy SOTA on LongBench.
The paper reveals that multimodal diffusion transformers often lose fine‑grained textual semantics in deeper layers—a phenomenon called Prompt Forgetting—and introduces Prompt Reinjection, a training‑free inference technique that re‑injects shallow text features to markedly improve text‑image alignment and instruction compliance while preserving visual quality and incurring negligible computational overhead.
The article showcases five ICML 2026 papers from the Taotian Group that tackle core multimodal AI challenges—interactive video try‑on, high‑resolution vision, e‑commerce video reasoning, sparse‑reward reinforcement learning, and curriculum learning for large language models—detailing their problem statements, novel solutions, and strong experimental results.
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.
