«Обживаемся» с Паспарту после переезда на Hermes и наступаем на новые «грабли». Ну и Codex подключаем — как же без него?

Продолжение истории про переезд ИИ-агента Паспарту с OpenClaw на Hermes. На этот раз — как я пересадил его на дешёвый DeepSeek, научил переключать модели на лету, подключил бесплатную NVIDIA Nemotron через OpenRouter и приделал Codex, который доводит сложные задачи до конца сам, пока ноутбук выключен. Грабли №16–25: стали мельче, но никуда не делись.

https://habr.com/ru/articles/1045892/

#Hermes #DeepSeek #OpenRouter #Nemotron #Codex #ИИагент #selfhosted #Telegramбот #OpenAI

#NVIDIA’s #Nemotron, the new genius on the block

Me:

You said: My training data associates Israel-related political content with antisemitism debates

Explain how is your training data doing that and why

It (truncated for clarity):

## How LLM Training Creates That Association

**Mechanism**: Statistical co-occurrence in the training corpus.

### What the training data actually contains

Vast amounts of text where "Israel" + "politics" + "antisemitism" appear together:

| Source type | Typical pattern |
|-------------|-----------------|
| News articles | "Israel-Hamas war sparks antisemitism debate on campuses" |
| Opinion columns | "Criticism of Israel vs. antisemitism: where's the line?" |
| Social media dumps | "Israel did X" → replies accusing/defending antisemitism |
| ADL/NGO reports | "Antisemitic incidents rise after Israel military actions" |
| Congressional hearings | "Antisemitism on campus: Israel-Palestine discourse" |
| Wikipedia | "Antisemitism in the 21st century: relationship to Israel" |
| Fact-checks | "Claim: criticizing Israel is antisemitic" |

### What the model learns

Not "Israel → antisemitism" as a fact. Rather: **P(antisemitism | Israel + politics + discourse) is high** in the training distribution.

So when I see:
- "Israel"
- "political conflict"
- "leaks"
- "administration"
- "media"

The conditional probability spikes for "antisemitism discourse" as an associated topic — because that cluster appears together *a lot* in the data.

[…]

This is a general LLM failure mode: **substituting distributional association for contextual judgment**. Happens with many sensitive topics (race, gender, religion, geopolitics).

#aicon #aihype #aibias #antisemitism

RT @PavloMolchanov: 🚀 Selbst-Spekulation ermöglicht eine 6,75-fache echte Beschleunigung der LLM-Generierung mit SGLang-Inference!

mehr auf Arint.info

#AI #Diffusion #LLM #MachineLearning #Nemotron #SGLang #arint_info

https://x.com/PavloMolchanov/status/2060245957254824246#m

NVIDIA Nemotron 3 Nano Omni: Open Multimodal AI Agent Guide 2026

NVIDIA released Nemotron 3 Nano Omni on April 28, 2026 — the first open model to natively unify vision, audio, and language in a shared reasoning loop, delivering 9x highe...

https://wowhow.cloud/blogs/nvidia-nemotron-3-nano-omni-multimodal-agent-developer-guide-2026

#wowhow #nvidia #nemotron #multimodalai

New week, more slides: Run LLMs Locally

Now with LFM 2 and new slides for using Transformers.js with WebGPU for Privacy Filter, Function Calling and Embeddings, running completely in your browser.

https://codeberg.org/thbley/talks/raw/branch/main/Run_LLMs_Locally_2026_ThomasBley.pdf

#ai #llm #llamacpp #stablediffusion #gptoss #qwen3 #glm #localai #gemma4 #nemotron #webgpu

RT @UnslothAI: NVIDIA veröffentlicht Nemotron-3-Nano-Omni, ein neues 30B offenes multimodales MoE-Modell. Nemotron-3-Nano-Omni-30B-A3B ist das stärkste Omni-Modell in seiner Größe und unterstützt Audio, Video, Bilder und Text. Läuft auf ~25GB RAM. GGUF: https://huggingface.co/unsloth/NVIDIA-Nemotron-3-Nano-Omni-30B-A3B-Reasoning-GGUF Anleitung: https://unsloth.ai/docs/models/nemotron-3-nano-omni

mehr auf Arint.info

#KI #MachineLearning #Multimodal #Nemotron #NVIDIA #OpenSource #arint_info

https://x.com/UnslothAI/status/2049161390150365344#m

RT @DJLougen: Das Nemotron-Omni SFT ist da! Nemostein-3-Nano-Omni-30b-a3b https://huggingface.co/GestaltLabs/Nemostein-3-Nano-Omni-30b-a3b GGUFs werden hochgeladen https://huggingface.co/GestaltLabs/Nemostein-3-Nano-Omni-30b-a3b-GGUF @NVIDIAAI ich hab dich verstanden

mehr auf Arint.info

#GGUF #HuggingFace #MachineLearning #Nemotron #NVIDIAAI #SFT #arint_info

https://x.com/DJLougen/status/2049531107092176976#m