Models & Labs
Hugging Face releases TRELLIS.2 fine-tuning guide
Hugging Face Bloghigh confidence
Why it matters
- →Enables precise geometric control in 3D generation by targeting specific model components with LoRA.
- →Prevents catastrophic forgetting of material priors by freezing the texture flow DiT during training.
- →Provides production-ready Docker containers and scripts, lowering the barrier to entry for custom 3D models.
Hugging Face has published a detailed guide and codebase for fine-tuning TRELLIS.2, a leading 3D generation model, using Low-Rank Adaptation (LoRA). The tutorial demonstrates how to improve geometric accuracy for specific object categories by training adapters on the sparse-structure and shape flow DiTs while keeping the texture flow frozen. Using a subset of the Google Scanned Objects dataset containing 188 footwear items, the team shows that this targeted approach corrects structural errors without degrading material quality. The release includes Docker containers for both preprocessing and training, making the workflow accessible via Hugging Face Jobs or local hardware.
Read originalThe story around this
Earlier coverage that leads up to this article, and what followed. Lines connect each piece to the closest one after it, converging here.
How we got here
- 1MIT Develops System to Enhance CAD Model Generation
MIT News AI · July 16, 2026 · Background
- 2NVIDIA and Hugging Face Enhance Diffusion Model Training
Hugging Face Blog · July 17, 2026 · Related
- 3Fine-tuning 350M Model for Structured Outputs
Hugging Face Blog · September 3, 2026 · Related
- 4InstructMesh repairs AI 3D models for real-world use
MIT News AI · October 1, 2026 · Background
- 5vLLM v0.31.0rc3 adds Model Runner V2 dummy inputs
vLLM Releases · October 2, 2026 · Background
- 6Hugging Face warns embedded AI evaluators lack standardization
TechCrunch AI · October 19, 2026 · Background
- 7llama.cpp aligns mtmd resize with torchvision
llama.cpp Releases · October 24, 2026 · Background
- 8llama.cpp decouples sparse attention from mask shape
llama.cpp Releases · October 28, 2026 · Background
More from Hugging Face Blog
© Hugging Face BlogModels & Labscoding
IBM releases LLMON markup to fix prompt injection
IBM is tackling the brittleness of small language models by introducing LLMON, a structured markup language that explicitly separates instructions from data. By using reserved special tokens instead of ambiguous characters like JSON brackets, these new Granite 4.1 preview models drastically reduce indirect prompt injection attacks while maintaining standard performance. This approach forces the model to bind specific constraints to specific data payloads, eliminating the confusion that often leads to hallucinations or security breaches in long contexts.
Hugging Face Blog
© Hugging Face BlogOpen Sourceagents
Tether AI launches BotPsy, a sub-billion agentic model
Tether AI Research has released BotPsy, an 0.8B-parameter model built on Qwen3.5 that specializes in tool-use and agentic workflows for on-device deployment. It achieves state-of-the-art performance in its size class, outperforming larger models like Qwen3.5-2B-Think on benchmarks such as BFCL v4 and ACEBench while maintaining strict privacy by keeping data local. The model is designed to handle routine tasks like calendar management and web search without relying on expensive cloud APIs, offering a viable path for private, self-hosted personal assistants. With open weights under Apache 2.0 and GGUF support, it provides immediate utility for developers building lightweight, autonomous agents.
Hugging Face Blog
© Hugging Face BlogResearchresearch
Pooling-Aware Training for Multi-Vector Embeddings
Multi-vector models like ColBERT are powerful but storage-heavy, often requiring hundreds of vectors per document. This technique teaches the encoder to preserve retrieval quality even when token vectors are aggressively merged via hierarchical clustering. By training on both full and pooled representations simultaneously, the model learns to compress information without losing critical distinctions. The result is a 4x reduction in vector count with only a marginal drop in accuracy, making large-scale semantic search significantly more efficient.
Hugging Face Blog
More in Models & Labs
Models & Labsother
llama.cpp b12340 adds CUDA 13 and ROCm 10.0
This release quietly closes the hardware gap for local inference by adding default support for CUDA 13 and ROCm 10.0 across Linux and Windows. Users with newer NVIDIA GPUs or AMD accelerators no longer need to compile from source to get performance, while Apple Silicon builds now ship KleidiAI disabled by default, shifting the focus back to standard ARM optimizations. The inclusion of Snapdragon support for Linux and Android marks a significant step toward on-device AI beyond the x86/ARM desktop ecosystem. It is less about new features and more about ensuring llama.cpp remains the universal runtime for every available accelerator.
llama.cpp Releases
Models & Labscoding
llama.cpp b12342 adds few-row MMQ for speculative decoding
Speculative decoding just got a significant speed boost on NVIDIA GPUs. This release introduces a specialized kernel for 'few-row' matrix multiplications, targeting the small batch sizes typical of verification passes in speculative decoding workflows. By optimizing these specific operations for quantized formats like Q8_0 and IQ4 variants, llama.cpp reduces latency where it matters most during token generation. The change is CUDA-only for now but leverages hardware capabilities from Ampere onwards, making local inference faster without requiring new model architectures or complex configuration changes.
llama.cpp Releases
© TechCrunch AIModels & Labsmodels
Google deploys Gemini 4 Hydrogen for faster Search
Google is quietly shifting its search infrastructure by replacing older fast models with Gemini 4 Hydrogen, a tier designed specifically for speed and cost-efficiency rather than raw reasoning power. This model completes tasks 2.5 times faster than its predecessor while beating the more powerful Argon on specific intelligence benchmarks, proving that specialized efficiency can outperform brute force in high-volume workflows. By integrating this into Search's fan-out technique, Google now breaks down complex queries into subqueries with greater precision and speed, delivering richer visual responses without the latency of heavier models. This marks a strategic pivot toward agentic execution layers within consumer search, making AI Mode feel snappier for everyday multi-part questions.
TechCrunch AI