AI Pulse by Inblix

AI Research

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

595 articles

Cohere's new RTEB benchmark catches models faking their retrieval scores — Inblix summary
Research

Cohere's new RTEB benchmark catches models faking their retrieval scores

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

Cohere just dropped a retrieval benchmark that's designed to embarrass models that have been gaming the system. The Ret...

I built the same game on Roblox, Unity, and the web. Only one didn't fall apart. — Inblix summary
Research

I built the same game on Roblox, Unity, and the web. Only one didn't fall apart.

Hugging Face Blog · Sep 29, 2025 · 2 min read

Andrej Karpathy's 'vibe coding' concept—using AI as a high-level programming language where you barely touch the code—s...

Intel Prunes Qwen3 Draft Model to Hit 1.4x Speedup on Local AI Agents — Inblix summary
Research

Intel Prunes Qwen3 Draft Model to Hit 1.4x Speedup on Local AI Agents

Hugging Face Blog · Sep 29, 2025 · 2 min read

Intel’s latest experiments are putting the squeeze on draft models to make local AI agents feel less like a waiting gam...

NVIDIA、600万件の合成ペルソナで「日本を理解するAI」の青写真を公開 — Inblix summary
Research

NVIDIA、600万件の合成ペルソナで「日本を理解するAI」の青写真を公開

Hugging Face Blog · Sep 26, 2025 · 1 min read

日本文化を真に理解するAIを構築する上で最大の障壁は、高品質な日本語データの絶対的な不足だった。NVIDIAはこの現実を正面から覆す。同社が発表したオープンデータセット「Nemotron-Personas-Japan」は、600万件の合成...

Swift Transformers Hits 1.0 With 100x Faster Jinja, Pivots to MLX and Agents — Inblix summary
Research

Swift Transformers Hits 1.0 With 100x Faster Jinja, Pivots to MLX and Agents

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

Two years after its quiet launch, the Swift Transformers library just hit version 1.0—and the team is signaling a major...

Hugging Face trains a 2.2B model to see GUIs, hitting 65% on ScreenSpot — Inblix summary
Research

Hugging Face trains a 2.2B model to see GUIs, hitting 65% on ScreenSpot

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

Turning a vision-language model with zero grounding ability into a GUI agent that can click and type is a monumental da...

ServiceNow drops SyGra, a free framework to stop your AI dataset nightmares — Inblix summary
Research

ServiceNow drops SyGra, a free framework to stop your AI dataset nightmares

Hugging Face Blog · Sep 22, 2025 · 2 min read

Let’s be honest: wrangling training data is the miserable, unglamorous part of AI work that nobody wants to talk about....

Meta's GAIA 2.0 benchmark humbles GPT-5 and Grok-4 with noisy, time-sensitive agent tasks — Inblix summary
Research

Meta's GAIA 2.0 benchmark humbles GPT-5 and Grok-4 with noisy, time-sensitive agent tasks

Hugging Face Blog · Sep 22, 2025 · 2 min read

The original GAIA benchmark is practically cooked. Two years after its 2023 release, the top models are acing the harde...

Scaleway Brings Sub-200ms AI Inference to Hugging Face, Starting at €0.20/M Tokens — Inblix summary
Research

Scaleway Brings Sub-200ms AI Inference to Hugging Face, Starting at €0.20/M Tokens

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

Hugging Face just got a serious European hardware upgrade. Scaleway is now a supported Inference Provider on the Hub, w...

Scanning 500K Hugging Face models found a 'dangerous middle' nobody talks about — Inblix summary
Research

Scanning 500K Hugging Face models found a 'dangerous middle' nobody talks about

Hugging Face Blog · Sep 18, 2025 · 2 min read

There are over half a million models on the Hugging Face hub, and until now, picking one felt like a gamble on safety....

Hugging Face adds Public AI as a free, nonprofit serverless inference provider — Inblix summary
Research

Hugging Face adds Public AI as a free, nonprofit serverless inference provider

Hugging Face Blog · Sep 17, 2025 · 2 min read

Hugging Face just made it significantly easier to run models from public-interest institutions without setting up your...

Hugging Face's LeRobot v3 packs millions of episodes into single files — Inblix summary
Research

Hugging Face's LeRobot v3 packs millions of episodes into single files

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

The Hugging Face robotics team just solved a scaling headache that's been quietly plaguing the community. LeRobotDatase...

Hugging Face makes AI watermarking a single-line command in Gradio — Inblix summary
Research

Hugging Face makes AI watermarking a single-line command in Gradio

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

Hugging Face just made it trivially easy to label synthetic content. The company rolled out a visible watermarking feat...

Writer drops 1.5B reasoning models that beat benchmarks, not your GPU — Inblix summary
Research

Writer drops 1.5B reasoning models that beat benchmarks, not your GPU

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

Writer just dropped a trio of tiny open models that punch well above their weight class. The Palmyra-mini family consis...

Hugging Face just made GPT-OSS's best tricks part of every model — Inblix summary
Research

Hugging Face just made GPT-OSS's best tricks part of every model

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

Remember how GPT-OSS got those impressive efficiency numbers? The secret wasn't just the architecture — it was a stack...

Together AI now lets you fine-tune any Hugging Face model in 5 minutes — Inblix summary
Research

Together AI now lets you fine-tune any Hugging Face model in 5 minutes

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

The gap between finding a useful model and actually making it your own has been a persistent headache. Together AI and...

Qwen3-4B Fine-Tune Jumps 15 Points on Data Science Tasks by Ditching Bloat — Inblix summary
Research

Qwen3-4B Fine-Tune Jumps 15 Points on Data Science Tasks by Ditching Bloat

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

If you’ve ever opened a massive, over-engineered agent framework and wondered if all that code is actually helping, the...

mmBERT cracks 1,800 languages to dethrone XLM-R as the first multilingual model to surge ahead — Inblix summary
Research

mmBERT cracks 1,800 languages to dethrone XLM-R as the first multilingual model to surge ahead

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

For years, XLM-R has been the untouchable gold standard for multilingual language models. Every new contender benchmark...

Google's 308M EmbeddingGemma is now the top multilingual model under 500M parameters — Inblix summary
Research

Google's 308M EmbeddingGemma is now the top multilingual model under 500M parameters

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

Google DeepMind dropped EmbeddingGemma today, and it's a genuinely practical release. We're talking about a 308-million...

SandboxAQ drops 5.2M AI-generated drug complexes, cuts pharma R&D time by 75% — Inblix summary
Research

SandboxAQ drops 5.2M AI-generated drug complexes, cuts pharma R&D time by 75%

Hugging Face Blog · Sep 2, 2025 · 2 min read

The biggest bottleneck in AI-driven drug discovery isn't algorithms. It's data — specifically, the lack of high-quality...

ZeroGPU cuts PyTorch cold starts by 1.8× with ahead-of-time compilation — Inblix summary
Research

ZeroGPU cuts PyTorch cold starts by 1.8× with ahead-of-time compilation

Hugging Face Blog · Sep 2, 2025 · 2 min read

The promise of ZeroGPU is simple and smart: stop paying for idle GPUs. Hugging Face Spaces grabs a GPU only when your m...

NVIDIA drops 6M multilingual reasoning examples, cutting edge model costs 60% — Inblix summary
Research

NVIDIA drops 6M multilingual reasoning examples, cutting edge model costs 60%

Hugging Face Blog · Aug 20, 2025 · 2 min read

NVIDIA isn't just releasing models anymore—it's open-sourcing the raw intelligence behind them. The company just publis...

Claude can now generate photorealistic images and perfect text with Krea and Qwen — Inblix summary
Research

Claude can now generate photorealistic images and perfect text with Krea and Qwen

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

Forget the plastic skin and garbled text that plague most AI images. Anthropic's Claude just got a major upgrade, and i...

Hugging Face's Kernel Builder Turns Solo CUDA Code Into Shared, Production-Ready Python Packages — Inblix summary
Research

Hugging Face's Kernel Builder Turns Solo CUDA Code Into Shared, Production-Ready Python Packages

Hugging Face Blog · Aug 18, 2025 · 2 min read

Writing a fast CUDA kernel for a specific GPU is one thing. Making it build cleanly for multiple architectures, survive...