Hugging Face – Posts

Join the conversation

Join the community of Machine Learners and AI enthusiasts.

All HF Hub posts

posted an update 2 days ago

Post

6862

Gemma 4 12B can now run locally on just 8GB RAM via Dynamic GGUFs.

Google's new model, Gemma 4 12B Unified supports image, audio and 256K context.
You can run and train the model via Unsloth Studio.

GGUF: unsloth/gemma-4-12b-it-GGUF
Guide: https://unsloth.ai/docs/models/gemma-4

AxionLab-official

posted an update 2 days ago

Post

6201

We're happy to announce that we released a Reasoning tuned version of Supra-50M!

SupraLabs/Supra-50M-Reasoning

hypothetical

posted an update 3 days ago

Post

3698

The smallest and the highest quality in the world Gemma4 E2B and E4B models! 7x compression! From 9.3GB -> 1.4GB!

TheStageAI/gemma-4-E2B-it
TheStageAI/gemma-4-E4B-it

2 replies

FlameF0X

posted an update 4 days ago

Post

6983

MiniMax-M3 coming soon.
https://github.com/MiniMax-AI/MiniMax-M3

evalstate

posted an update 3 days ago

Post

3163

Hugging Face MCP Server v0.3.17
~~~~~~~~~~~~~~~~~~~~~~~~~~~~

SEP-2640 "Skills Over MCP" support added (early access)

2 replies

ovi054

posted an update 3 days ago

Post

2538

Qwen Image Edit 2511 Fast + LoRA ⚡

ovi054/Qwen-Image-Edit-2511-LoRA

QIE-2511 is an image editing model with integrated LoRA capabilities. You can add any custom LoRA to generate and edit images within this Space.

👉 Try it now: ovi054/Qwen-Image-Edit-2511-LoRA

RiverRider

posted an update 1 day ago

Post

Words do not have determined meanings.

The vocabulary itself is reflexive. It is self-referential, looping back into its own structure rather than anchoring in fixed reality. What we treat as stable meaning is continually reconstituted in the act of using it. The observers own interpretations molding each word like clay with every utterance.

All large language models to date treat words otherwise. At the moment of softmax crystallization they determine the meaning of every token. Probabilities collapse into a single output. Meaning is not found. It is fixed, token by token, in that final distribution.

SRT-Introspect is a demo for observing what Qwen actually thinks at the points of highest effort. It surfaces the internal representations during generation, making visible the reflexive vocabulary at work and the precise crystallization process: the weights, the assumptions, the decisions that resolve ambiguity into output. This includes accounting for anisotropy collapse in hidden states by centering representations around the layer-mean before analysis.

Feel free to comment your prompts

RiverRider/srt-introspect

Repo
https://github.com/space-bacon/SRT

1 reply

kanaria007

posted an update 1 day ago

Post

✅ Article highlight: *Revocable Releases, Subject Scopes, and Unlearning Verification for Learning Worlds* (art-60-173, v0.1)

TL;DR:
This article argues that once you release data, forgetting becomes a supply-chain problem.

A world can promise future exclusion, controlled-channel revocation, or bounded unlearning claims—but only if those claims are receipted. To say “Release R is revocable,” “Subject X was forgotten,” or “Model M unlearned X,” you need pinned release contracts, precise subject scopes, scope-resolution receipts, and verification packs. Otherwise you are just telling a comforting story.

Read:
kanaria007/agi-structural-intelligence-protocols

Why it matters:
• turns “forgetting” into a governed lifecycle rather than a vague promise
• separates revocable releases from irreversible public redistribution
• makes “Subject X” precise enough to be caseable and auditable
• forces unlearning claims to be tested, bounded, and published honestly

What’s inside:
• *release contracts* with revocation tiers and downstream obligations
• *subject selector* + *scope resolution* artifacts for “where X might exist”
• *unlearning contracts* and *verification packs* for testable forgetting claims
• explicit irreversibility disclosures, so public claims do not promise impossible erasure
• bounded public claim shapes under publication policy

Key idea:
Do not say:

*“we forgot X.”*

Say:

*“this release had this revocation tier, this subject scope was resolved across corpora/releases/models, this unlearning execution and verification pack were run, and these are the limits of what we can and cannot guarantee.”*

Shrijanagain

posted an update 1 day ago

Post

Excited to launch SKT-ST-X-0-3B by SKT AI Labs! 🚀🇮🇳

A powerful 3B Parameter Mixture of Experts (MoE) model optimized for high-performance reasoning with a small footprint.

--> Quick Specs:
> Total Params: ~3B | Active Params: ~1.1B (2 experts/token)
> Pre-trained on 40B tokens (SKT-OMNI-CORPUS-2T)

1.Context: 8K tokens
2.Bilingual: English & Hindi 🇬🇧🇮🇳
3. Base: Built on ST-X-0 with Mixtral stability

Get 3B intelligence at 1B inference speeds. Fully open-source under Apache-2.0! 👇

🔗 Try it on Hugging Face: sKT-Ai-Labs/SKT-ST-X-0-3B

#AI #OpenSource #LLM #MixtureOfExperts #SKTAILabs #MachineLearning

pedrodev2026

posted an update 2 days ago

Post

102

O NanoMathDataset é um dataset de quase 4M linhas de contas de matemática básica, veja ele em: pedrodev2026/NanoMathDataset

Recently active users