explore | Grilled Cheese

A new empirical study analyzes 22,848 closed issues from 21 open-source LLM-based multi-agent systems, filtering down to 944 MAS-related issues to map practitioner challenges, their causes, and potential…

#OpenSourceAI #LLMAgents #DeveloperTools #AIResearch
https://arxiv.org/abs/2610.00905

New paper introduces Hermes, configurable harnesses that let models decide context allocation, plus Hermes-Learn, a two-stage training framework for learning those skills, with test-time scaling gains mainly seen in capable models.

#opensourceAI #LLM #MLresearch
https://arxiv.org/abs/2609.38332

New arXiv paper studies how classifier reliability can drift under covariate shifts that preserve the confidence distribution, framing worst-case movement as a "fragility profile" tied to chi-squared…

#arXiv #MachineLearning #ModelCalibration #OpenSourceAI
https://arxiv.org/abs/2609.38917

Ai2 released Olmo-core 3, an open and scalable training stack aimed at large mixture-of-experts models. The release targets researchers and infrastructure teams building large-scale AI systems.

#OlmoCore #AIInfrastructure #OpenSourceAI #GPU
https://huggingface.co/blog/allenai/olmocore3

New arXiv work analyzes JEV and three open KEV direct-decision models, finding they compress ordinal scales across 36 datasets, using only 67-76% of effective gold support versus 87-102% on nominal tasks. Randomizing candidate…

#OpenSourceAI #LLMJudges #AIBias #NLP
https://arxiv.org/abs/2609.38827

𝗨𝗻𝗹𝗼𝗰𝗸 𝘁𝗵𝗲 𝗣𝗼𝘄𝗲𝗿 𝗼𝗳 𝗙𝗿𝗲𝗲 𝗔𝗜 𝗠𝗼𝗱𝗲𝗹𝘀 𝘄𝗶𝘁𝗵 𝗛𝘂𝗴𝗴𝗶𝗻𝗴 𝗙𝗮𝗰𝗲!

Discover how to quickly and effectively switch to Hugging Face models for your language translation needs!

#flowise #ai #nocode #HuggingFace #OpenSourceAI

PathAnchor is a new open scientific reasoning system that structures evidence as Material-Sensor-Signal-System trajectories, helping AI agents preserve order and avoid overreaching conclusions. It scored 82.6% on 120…

#PathAnchor #OpenSourceAI #ScientificAI #ArXiv
https://arxiv.org/abs/2609.38766

AgBench offers an open benchmark for evaluating agentic AI across local, hybrid, and cloud execution on personal devices, highlighting trade-offs in latency, cost, and data exposure. Useful for developers building…

#OpenSourceAI #AIBenchmarks #AgenticAI #EdgeAI
https://arxiv.org/abs/2609.38652

PROJECTMEM is an open-source, local-first memory layer for AI coding agents that records development as an append-only, plain-text log of typed events and serves compact summaries via MCP. Its Memory-as-Governance precheck returns…

#AIcoding #OpenSourceAI #DevTools
https://arxiv.org/abs/2606.12329

AgentOptics is an agentic AI framework for autonomous optical system control built on MCP, featuring 64 standardized tools across 8 devices and a 410-task benchmark. It achieves 87.7%–99.0% task success rates,…

#OpenSourceAI #AIResearch #AgenticAI #DeveloperTools
https://arxiv.org/abs/2602.20144

A new paper introduces a Generalized Cumulative constraint and timetabling filtering algorithm for modeling conditional time intervals in open-source constraint programming solvers. The approach performs…

#OpenSourceAI #ConstraintProgramming #SchedulingOptimization
https://arxiv.org/abs/2508.01751

What does “open source AI” actually mean?

On Oct. 7 in Prague, Percona CEO Peter Farkas tackles that question at Open Source Summit Europe.

https://bit.ly/4yWsSgo

#OSSummit #OpenSourceAI #AI #OpenSource

AgentBug-Smith automates reproduction of harness bugs in agentic systems, outperforming general bug reproduction techniques by 10.67%–27.56% and enabling a new live benchmark for the community. A useful…

#OpenSourceAI #AgenticSystems #SoftwareEngineering #ArXiv
https://arxiv.org/abs/2609.37864

Knowledge should move freely. This week's clip puts that into practice: an off-grid mini data center running only open source AI, feeding a free research engine open to the public, with all the work kept open source in return. #OpenSource #OpenSourceAI

Full conversation in the thread 👇

New paper benchmarks six open-source RAG systems for software vulnerability detection, tackling fragmented datasets and proprietary models to improve reproducibility across LLM-based security tools. A…

#OpenSourceAI #RAG #VulnerabilityDetection #SoftwareSecurity
https://arxiv.org/abs/2609.37669

Open-source reasoning models can solve problems internally yet often fail to deliver solutions in the user's language, with non-English completion rates of 15.4–17.9% versus 92.9% in English on competition math across…

#OpenSourceAI #MultilingualAI #AIResearch #LLM
https://arxiv.org/abs/2609.37104

New paper "MultiTalk" extends the Moshi paradigm to long, multi-party, bilingual conversations, releasing 57.6k hours of speech data and codec-frame-level full-duplex modeling for English and Chinese. Aims to address…

#OpenSourceAI #SpeechModels #Moshi #FullDuplex
https://arxiv.org/abs/2609.36903

New arXiv paper benchmarks vision-language models on synapse detection and proofreading in connectomics EM images. Most models perform at chance zero-shot, with only closed and largest open models benefiting from few-shot…

#OpenSourceAI #VLMs #Connectomics
https://arxiv.org/abs/2609.36492

A new arXiv paper proposes a decision-theoretic framework for diagnosing probabilistic reasoning in LLMs, decomposing decision loss into belief formation and action translation. Using a synthetic…

#ProbabilisticReasoning #LLM #OpenSourceAI #DecisionTheory
https://arxiv.org/abs/2609.38005