AI Pulse by Inblix

AI Research

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

595 articles

PipelineRL ditches value functions, matches Open-Reasoner-Zero with inflight weight updates — Inblix summary
Research

PipelineRL ditches value functions, matches Open-Reasoner-Zero with inflight weight updates

Hugging Face Blog · Apr 25, 2025 · 2 min read

The team behind PipelineRL just dropped a compelling rebuttal to one of reinforcement learning's most stubborn trade-of...

Hugging Face's Tiny Agent: MCP strips agents down to a 50-line while loop — Inblix summary
Research

Hugging Face's Tiny Agent: MCP strips agents down to a 50-line while loop

Hugging Face Blog · Apr 25, 2025 · 2 min read

The hype around the Model Context Protocol (MCP) has been deafening, but after weeks of digging into it, the reality is...

olmOCR fine-tune fixes the header-footer blind spot that killed it for invoices — Inblix summary
Research

olmOCR fine-tune fixes the header-footer blind spot that killed it for invoices

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

Pipeline-based OCR engines have a dirty secret: they often butcher the reading order, especially on layout-rich pages....

TNG runs 5,000+ LLM inferences per hour on 24 H100s—here's the prefill/decode trade-off — Inblix summary
Research

TNG runs 5,000+ LLM inferences per hour on 24 H100s—here's the prefill/decode trade-off

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

Most teams treat LLM inference as a black box. TNG doesn't have that luxury. They're running over 5,000 inferences per...

Cohere now serves its own enterprise models directly on Hugging Face Hub — Inblix summary
Research

Cohere now serves its own enterprise models directly on Hugging Face Hub

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

Hugging Face just landed a first for its Inference Providers program: Cohere is not only joining the roster, but it's t...

Gradio Isn’t Just UI — Its API Recorder and SSR Kill the Frontend/Backend Split — Inblix summary
Research

Gradio Isn’t Just UI — Its API Recorder and SSR Kill the Frontend/Backend Split

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

Calling Gradio a UI library is like calling a smartphone a camera. It’s technically true, but it misses the entire oper...

Princeton’s HELMET benchmark exposes synthetic AI tests as useless for real work — Inblix summary
Research

Princeton’s HELMET benchmark exposes synthetic AI tests as useless for real work

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

The team at Princeton NLP just dropped a new evaluation suite called HELMET at ICLR 2025, and its core finding is blunt...

Hugging Face acquires Pollen Robotics, will sell the $70,000 open-source Reachy 2 — Inblix summary
Research

Hugging Face acquires Pollen Robotics, will sell the $70,000 open-source Reachy 2

Hugging Face Blog · Apr 14, 2025 · 2 min read

Hugging Face just made its most direct bet yet that AI's next big platform shift is physical. The company has acquired...

Hugging Face scanned 4.4M models: 51,700 repos flagged with hidden code risks — Inblix summary
Research

Hugging Face scanned 4.4M models: 51,700 repos flagged with hidden code risks

Hugging Face Blog · Apr 14, 2025 · 2 min read

Six months into a security partnership, Protect AI has scanned 4.47 million unique model versions across 1.41 million r...

Salamandra 7B gains vision via SigLIP, trained on 6.1M multilingual samples — Inblix summary
Research

Salamandra 7B gains vision via SigLIP, trained on 6.1M multilingual samples

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

The Language Technologies Lab has given eyes to its Salamandra 7B model, unveiling Visual Salamandra — a multimodal sys...

Hugging Face's FastRTC taps Cloudflare's 335-location TURN network for free global AI voice — Inblix summary
Research

Hugging Face's FastRTC taps Cloudflare's 335-location TURN network for free global AI voice

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

The absolute worst part of building real-time AI voice apps isn't the model — it's the networking. WebRTC, the protocol...

Arabic AI models flunk harder benchmarks: Top score plunges from 83% to 70% — Inblix summary
Research

Arabic AI models flunk harder benchmarks: Top score plunges from 83% to 70%

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

The rug got pulled out from under Arabic large language models this week. MBZUAI and the Arabic-Leaderboards team tight...

Meta drops Llama 4: 10M context window and a bold bet on NoPE layers — Inblix summary
Research

Meta drops Llama 4: 10M context window and a bold bet on NoPE layers

Hugging Face Blog · Apr 5, 2025 · 2 min read

Forget the slow drip of model releases. Meta just dumped two Llama 4 variants on Hugging Face—Maverick and Scout—and th...

Gradio hit 1M monthly devs by ignoring the 'productivity illusion' of high-level code — Inblix summary
Research

Gradio hit 1M monthly devs by ignoring the 'productivity illusion' of high-level code

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

When Gradio launched five years ago, it was a simple tool to let Stanford researchers slap a web interface on a compute...

Hugging Face rebrands its 3-year-old NLP course as 'The LLM Course' — Inblix summary
Research

Hugging Face rebrands its 3-year-old NLP course as 'The LLM Course'

Hugging Face Blog · Apr 3, 2025 · 2 min read

Hugging Face is officially retiring the 'NLP Course' branding that has been a cornerstone of open-source AI education f...

How TNG Broke vLLM’s FIFO Queue to Stop Power Users from Hogging GPUs — Inblix summary
Research

How TNG Broke vLLM’s FIFO Queue to Stop Power Users from Hogging GPUs

Hugging Face Blog · Apr 2, 2025 · 3 min read

If you’ve ever waited for a GPU while a single user floods an inference engine with hundreds of requests, you know the...

Hugging Face ditched .env files and HashiCorp Vault to stop secret sprawl after a breach — Inblix summary
Research

Hugging Face ditched .env files and HashiCorp Vault to stop secret sprawl after a breach

Hugging Face Blog · Mar 31, 2025 · 2 min read

When a security incident exposed credentials at Hugging Face, the team knew its patchwork of .env files and AWS-specifi...

Hugging Face ends its Gaudi fork, baking Intel AI chip support directly into TGI — Inblix summary
Research

Hugging Face ends its Gaudi fork, baking Intel AI chip support directly into TGI

Hugging Face Blog · Mar 28, 2025 · 2 min read

Hugging Face just made life a lot easier for anyone deploying large language models on Intel's Gaudi hardware. The team...

DeepSeek drops V3-0324 with MIT license and a 19.8-point math leap — Inblix summary
Research

DeepSeek drops V3-0324 with MIT license and a 19.8-point math leap

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

The newest model from DeepSeek isn't a flashy launch. It's a silent update that speaks volumes. DeepSeek-V3-0324, an up...

A tiny finetuned reranker just crushed public models 4x its size — Inblix summary
Research

A tiny finetuned reranker just crushed public models 4x its size

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

You don't need a massive, general-purpose reranker to get state-of-the-art search results. A new technical deep-dive an...

Gradio closes 70+ dataframe issues, finally adds multi-cell selection and search — Inblix summary
Research

Gradio closes 70+ dataframe issues, finally adds multi-cell selection and search

Hugging Face Blog · Mar 24, 2025 · 2 min read

If you've built anything with Gradio in the last two years, you've probably used gr.Dataframe. It's the workhorse for l...

Hugging Face Overhauls Inference Analytics Dashboard After Its Own Engineers Hit a Wall — Inblix summary
Research

Hugging Face Overhauls Inference Analytics Dashboard After Its Own Engineers Hit a Wall

Hugging Face Blog · Mar 21, 2025 · 2 min read

If you've ever stared at a sluggish analytics dashboard while your models were quietly melting down, Hugging Face just...

OlympicCoder 7B, a tiny open model, just beat GPT-4o and Claude 3.7 at live coding — Inblix summary
Research

OlympicCoder 7B, a tiny open model, just beat GPT-4o and Claude 3.7 at live coding

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

The open-source model OlympicCoder 7B is now topping the LiveCodeBench leaderboard, outperforming the daily-driver codi...

Hugging Face to White House: Open 7B models now beat Claude 3.7 on code — Inblix summary
Research

Hugging Face to White House: Open 7B models now beat Claude 3.7 on code

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

Hugging Face just told the White House something that should make proprietary AI vendors nervous: a 7-billion-parameter...