Learning to Communicate: Toward End-to-End Optimization of Multi-Agent Language Systems
DiffMAS framework enables joint optimization of latent communication protocols in multi-agent LLM systems via differentiable training.
Search the full wire by company, model, lab, or keyword. Every story we have ever aggregated.
DiffMAS framework enables joint optimization of latent communication protocols in multi-agent LLM systems via differentiable training.
Logic-based temporal event detection system infers high-level medical events from timestamped clinical data and background rules.
Mechanism design framework addresses decentralized risk analytics and compliance moral hazard in banking AML networks via temporal value assignment.
SemEval-2026 Task 4 introduces NSNRL benchmark for narrative similarity classification and embedding representation evaluation on 1000+ story triples.
Moving somewhere else next billing cycle. Two hours of coding on Max and I'm capped. Whatever you changed, change it back or say something. Silent nerfs to a paid product are a bad look.
Automatic status alert: Opus 4.7 experienced elevated error rates on 2026-04-23.
LLM leaderboards are widely used to compare models and guide deployment decisions. However, leaderboard rankings are shaped by evaluation priorities set by benchmark designers, rather than by the diverse goals and constraints of actual users and organizations. A single…
Online misinformation is one of the most challenging issues lately, yielding severe consequences, including political polarization, attacks on democracy, and public health risks. Misinformation manifests in any platform with a large user base, including online social networks…
Existing audio question answering benchmarks largely emphasize sound event classification or caption-grounded queries, often enabling models to succeed through shortcut strategies, short-duration cues, lexical priors, dataset-specific biases, or even bypassing audio via…
Data is a central resource for modern enterprises, and data validation is essential for ensuring the reliability of downstream applications. However, existing automated data unit testing frameworks are largely task-agnostic: they validate datasets without considering the…
Reasoning LLMs often spend substantial tokens on long intermediate reasoning traces (e.g., chain-of-thought) when solving new problems. We propose to summarize and store reusable reasoning skills distilled from extensive deliberation and trial-and-error exploration, and to…
Physics-informed neural networks (PINNs) have garnered significant interest for their potential in solving partial differential equations (PDEs) that govern a wide range of physical phenomena. By incorporating physical laws into the learning process, PINN models have…
Research on interpretable facial dynamics for deepfake detection using bio-behavioral features and traditional ML classifiers, with human perceptual validation.
GPT-5.5 and GPT-5.5 Pro variants detected on OpenRouter API platform with dated versions.
CRNN surrogate models for crystal growth dynamics simulation with explicit and implicit supersaturation conditioning architectures.
Study reveals LLM cultural biases show preference toward Japanese and non-Western content via Culture-Related Open Questions dataset.
StructMem proposes hierarchical memory framework for LLM agents balancing relational structure preservation with efficiency for long-horizon reasoning.
Plants from OpenAI, Meta, xAI, and Microsoft could emit more than 129M tons annually.
Proposes GROUNDING.md field-scoped epistemic grounding document to improve agentic AI coding reliability in scientific domains.
Mobile image enhancement model with gated encoding and multi-scale refinement bridging training-deployment gap for quantized inference.
Unsubstantiated post with no context or verifiable claim.
Mixed-methods study on university students' psychological willingness to disclose AI use via Cognition-Affect-Conation framework.
Ramen framework enables test-time adaptation of vision-language models under mixed-domain distribution shifts via active sample selection.
Agent Evolving Learning framework enables LLM agents to accumulate and leverage experience across open-ended episodes via two-timescale Thompson sampling.
X-GRAM proposes frequency-aware dynamic token injection and hybrid hashing to improve embedding parameter efficiency for large lookup tables.
Code LLMs generate ML pipelines with 87.7% inclusion of sensitive attributes in feature selection, revealing underestimated bias beyond simple conditional statements.
Framework for fairness in sequential decision-making under uncertainty via counterfactual inference, addressing online ML applications with cascading decisions.
Tencent releases Hy3 preview: 295B MoE model with 21B active parameters on Hugging Face.
Cross-lingual dysarthria severity assessment across 3,374 speakers from 12 languages shows aetiology-specific phonological degradation patterns in SSL speech models.
Technical walkthrough: Qwen 3.6 27B achieves 85 TPS, 125K context on single RTX 3090 using llama.cpp.