AI Pulse by Inblix

AI Research

Cutting-edge AI research papers, breakthroughs, and academic developments.

595 articles

Hugging Face silently moves 500,000+ repos off Git LFS with near-zero complaints — Inblix summary
Research

Hugging Face silently moves 500,000+ repos off Git LFS with near-zero complaints

Hugging Face Blog · Jul 15, 2025 · 2 min read

Hugging Face has pulled off what might be the quietest massive infrastructure migration in developer tools, transitioni...

Kimina-Prover hits 92.2% on math benchmark by learning to reuse its own lemmas — Inblix summary
Research

Kimina-Prover hits 92.2% on math benchmark by learning to reuse its own lemmas

Hugging Face Blog · Jul 10, 2025 · 2 min read

The team at Numina and Kimi just pushed automated theorem proving past a major milestone. Their new Kimina-Prover-72B s...

Hugging Face deploys MCP server with Streamable HTTP, dodging stateful headaches for 1M+ models — Inblix summary
Research

Hugging Face deploys MCP server with Streamable HTTP, dodging stateful headaches for 1M+ models

Hugging Face Blog · Jul 10, 2025 · 2 min read

Hugging Face just detailed the architecture behind its official Model Context Protocol (MCP) server, and the big takeaw...

Hugging Face’s async robot brains cut task time in half by never waiting to think — Inblix summary
Research

Hugging Face’s async robot brains cut task time in half by never waiting to think

Hugging Face Blog · Jul 10, 2025 · 2 min read

The dirty secret of most robot learning models isn't their failure rate—it's how much time they spend just sitting ther...

ScreenEnv drops a 10-second Docker desktop sandbox for AI agents — Inblix summary
Research

ScreenEnv drops a 10-second Docker desktop sandbox for AI agents

Hugging Face Blog · Jul 10, 2025 · 2 min read

Desktop automation for AI agents has been a mess of brittle scripts and heavyweight VM setups. ScreenEnv changes that c...

Custom CUDA kernels juice Llama 3.1 405B on AMD MI300X, skipping Nvidia lock-in — Inblix summary
Research

Custom CUDA kernels juice Llama 3.1 405B on AMD MI300X, skipping Nvidia lock-in

Hugging Face Blog · Jul 9, 2025 · 2 min read

A billion-plus requests a day. That's the ballpark for ChatGPT, and it underscores why even a 1% latency shave in large...

Hugging Face Spaces just became the app store for LLMs with MCP support — Inblix summary
Research

Hugging Face Spaces just became the app store for LLMs with MCP support

Hugging Face Blog · Jul 9, 2025 · 2 min read

Hugging Face Spaces has quietly become the most interesting thing to happen to LLM tools since function calling. With G...

This $399 open-source robot from Hugging Face and Pollen Robotics ships as a kit you build yourself — Inblix summary
Research

This $399 open-source robot from Hugging Face and Pollen Robotics ships as a kit you build yourself

Hugging Face Blog · Jul 9, 2025 · 2 min read

Pollen Robotics and Hugging Face just dropped Reachy Mini, a $399 desktop robot kit built squarely for the tinkerers, t...

Hugging Face's 3 silent alerts that keep model infrastructure from bleeding cash — Inblix summary
Research

Hugging Face's 3 silent alerts that keep model infrastructure from bleeding cash

Hugging Face Blog · Jul 8, 2025 · 2 min read

Running a platform that serves thousands of machine learning models means your cloud bill can go from predictable to ho...

Hugging Face's SmolLM3 packs 4B-level reasoning into a 3B model you can actually run — Inblix summary
Research

Hugging Face's SmolLM3 packs 4B-level reasoning into a 3B model you can actually run

Hugging Face Blog · Jul 8, 2025 · 2 min read

Hugging Face just dropped SmolLM3, and the numbers don't lie: a 3-billion-parameter model that outperforms Meta's Llama...

That gray in your training batch is burning cash: How a knapsack cut padding waste from 60% to zero — Inblix summary
Research

That gray in your training batch is burning cash: How a knapsack cut padding waste from 60% to zero

Hugging Face Blog · Jul 8, 2025 · 2 min read

If you've trained a multimodal model, you know the pain: GPUs sitting idle while your data loader catches up, and batch...

NeurIPS 2025 bets $16K that benchmarks, not loss curves, can judge baby LLMs — Inblix summary
Research

NeurIPS 2025 bets $16K that benchmarks, not loss curves, can judge baby LLMs

Hugging Face Blog · Jul 4, 2025 · 2 min read

The blank stare of a loss curve is a familiar frustration for anyone who has trained a large language model. In the ear...

Hugging Face Ships Full Toolkit to Train Sparse Embedding Models Like SPLADE — Inblix summary
Research

Hugging Face Ships Full Toolkit to Train Sparse Embedding Models Like SPLADE

Hugging Face Blog · Jul 1, 2025 · 2 min read

Hugging Face just published a comprehensive guide for finetuning sparse embedding models using their Sentence Transform...

NVIDIA drops Llama Nemotron 8B VLM to dethrone OCR leaders on Hugging Face — Inblix summary
Research

NVIDIA drops Llama Nemotron 8B VLM to dethrone OCR leaders on Hugging Face

Hugging Face Blog · Jun 27, 2025 · 3 min read

NVIDIA just dropped a focused bomb on the Vision Language Model space, and it’s aimed squarely at the unglamorous but m...

Google's Gemma 3n Crams 5B-Parameter Brains Into Just 2GB of VRAM — Inblix summary
Research

Google's Gemma 3n Crams 5B-Parameter Brains Into Just 2GB of VRAM

Hugging Face Blog · Jun 26, 2025 · 2 min read

Google just dropped a pair of models that flip the script on what “small” AI means. The Gemma 3n series, now fully inte...

SGLang taps Hugging Face transformers as a backend, instantly unlocking models like Kyutai's Helium — Inblix summary
Research

SGLang taps Hugging Face transformers as a backend, instantly unlocking models like Kyutai's Helium

Hugging Face Blog · Jun 23, 2025 · 2 min read

SGLang, the inference engine built for high-throughput and low-latency AI, just closed a major gap in its ecosystem. It...

Fine-tune FLUX.1-dev on an RTX 4090: QLoRA slashes VRAM below 10GB — Inblix summary
Research

Fine-tune FLUX.1-dev on an RTX 4090: QLoRA slashes VRAM below 10GB

Hugging Face Blog · Jun 19, 2025 · 2 min read

For all the staggering image generation capabilities of models like black-forest-labs/FLUX.1-dev, the hardware wall for...

Groq's LPU chips land on Hugging Face, bringing sub-100ms inference to open models — Inblix summary
Research

Groq's LPU chips land on Hugging Face, bringing sub-100ms inference to open models

Hugging Face Blog · Jun 16, 2025 · 2 min read

The wait for truly fast, on-demand open-source AI just got shorter. Groq, the inference provider that famously eschews...

vLLM's long prompts silently choke fast token generation for everyone — Inblix summary
Research

vLLM's long prompts silently choke fast token generation for everyone

Hugging Face Blog · Jun 12, 2025 · 2 min read

Here's a headache every LLM inference engineer knows too well: a single user with a massive prompt can bring response t...

Hugging Face adds Featherless AI for serverless access to models like DeepSeek-R1 — Inblix summary
Research

Hugging Face adds Featherless AI for serverless access to models like DeepSeek-R1

Hugging Face Blog · Jun 12, 2025 · 2 min read

Hugging Face just plugged a major gap in its inference ecosystem by onboarding Featherless AI as a supported provider....

Hugging Face Kernel Hub erases 96 GB and hours of build pain with a single import — Inblix summary
Research

Hugging Face Kernel Hub erases 96 GB and hours of build pain with a single import

Hugging Face Blog · Jun 12, 2025 · 2 min read

Nobody gets into machine learning because they love babysitting a CUDA compiler. But until now, leveraging custom-optim...

NVIDIA's GR00T N1.5 robot model now fine-tunes on a $200 open-source arm — Inblix summary
Research

NVIDIA's GR00T N1.5 robot model now fine-tunes on a $200 open-source arm

Hugging Face Blog · Jun 11, 2025 · 2 min read

NVIDIA just lowered the barrier to entry for generalist robot AI in a big way. The company announced Isaac GR00T N1.5,...

Hugging Face and NVIDIA team up to rent GPU clusters by the training run — Inblix summary
Research

Hugging Face and NVIDIA team up to rent GPU clusters by the training run

Hugging Face Blog · Jun 11, 2025 · 2 min read

The compute gap between AI's haves and have-nots has looked like a widening chasm lately, with gigawatt-scale superclus...

Hugging Face's ScreenSuite stress-tests 5 AI models on pure vision — no DOM cheating — Inblix summary
Research

Hugging Face's ScreenSuite stress-tests 5 AI models on pure vision — no DOM cheating

Hugging Face Blog · Jun 6, 2025 · 2 min read

Evaluating AI that can actually use a computer like a person — by looking at the screen — remains surprisingly hard to...