HuggingFace introduced ml-intern, the agent that just automated the post-training team
It's an open-source implementation of the real research loop that ML researchers do every day.
You give it a prompt, it researches papers, goes through citations, implements ideas in GPU sandboxes, iterates and builds deeply research-backed models for any use case. All built on the Hugging Face ecosystem.
How it works?
ml-intern makes full use of the HF ecosystem:
- finds papers on arxiv and hf.co/papers, reads them fully, walks citation graphs, pulls datasets referenced in methodology sections and on hf.co/datasets
- browses the Hub, reads recent docs, inspects datasets and reformats them before training so it doesn't waste GPU hours on bad data
- launches training jobs on HF Jobs if no local GPUs are available, monitors runs, reads its own eval outputs, diagnoses failures, retrains
ml-intern deeply embodies how researchers work and think. It knows how data should look like and what good models feel like.
CLI
Web + mobile
Also provisioned 1k$ GPU resources and Anthropic credits for you to use.
It's an open-source implementation of the real research loop that ML researchers do every day.
You give it a prompt, it researches papers, goes through citations, implements ideas in GPU sandboxes, iterates and builds deeply research-backed models for any use case. All built on the Hugging Face ecosystem.
How it works?
ml-intern makes full use of the HF ecosystem:
- finds papers on arxiv and hf.co/papers, reads them fully, walks citation graphs, pulls datasets referenced in methodology sections and on hf.co/datasets
- browses the Hub, reads recent docs, inspects datasets and reformats them before training so it doesn't waste GPU hours on bad data
- launches training jobs on HF Jobs if no local GPUs are available, monitors runs, reads its own eval outputs, diagnoses failures, retrains
ml-intern deeply embodies how researchers work and think. It knows how data should look like and what good models feel like.
CLI
Web + mobile
Also provisioned 1k$ GPU resources and Anthropic credits for you to use.
huggingface.co
Daily Papers - Hugging Face
Your daily dose of AI research from AK
🆒5
This is cool. Meet Simula from Google
A Google framework produces better synthetic datasets.
Instead of just randomly generating data or copying existing examples, Simula uses AI to strategically architect an entire dataset from scratch.
A Google framework produces better synthetic datasets.
Instead of just randomly generating data or copying existing examples, Simula uses AI to strategically architect an entire dataset from scratch.
Google Research
Designing synthetic datasets for the real world: Mechanism design and reasoning from first principles
To address the scarcity of data required for specialized AI, we introduce Simula, a framework that reframes synthetic data generation as dataset-level mechanism design. By using reasoning to architect datasets from first principles, Simula enables fine-grained…
❤4
Anthropic's Mythos AI model is being accessed by unauthorized users 🤣
A handful of users in a private online forum gained access to Mythos on the same day that Anthropic first announced a plan to release the model to a limited number of companies for testing purposes, said the person, who asked not to be named for fear of reprisal. The group has been using Mythos regularly since then, though not for cybersecurity purposes, said the person, who corroborated the account with screenshots and a live demonstration of the model.
The users relied on a mix of tactics to get into Mythos. These included using access the person had as a worker at a third-party contractor for Anthropic and trying commonly used internet sleuthing tools often employed by cybersecurity researchers, the person said. The users are part of a private Discord channel that focuses on hunting for information about unreleased models, including by using bots to scour for details that Anthropic and others have posted on unsecured websites such as GitHub.
“We’re investigating a report claiming unauthorized access to Claude Mythos Preview through one of our third-party vendor environments,” a spokesperson for Anthropic said in a statement.
A handful of users in a private online forum gained access to Mythos on the same day that Anthropic first announced a plan to release the model to a limited number of companies for testing purposes, said the person, who asked not to be named for fear of reprisal. The group has been using Mythos regularly since then, though not for cybersecurity purposes, said the person, who corroborated the account with screenshots and a live demonstration of the model.
The users relied on a mix of tactics to get into Mythos. These included using access the person had as a worker at a third-party contractor for Anthropic and trying commonly used internet sleuthing tools often employed by cybersecurity researchers, the person said. The users are part of a private Discord channel that focuses on hunting for information about unreleased models, including by using bots to scour for details that Anthropic and others have posted on unsecured websites such as GitHub.
“We’re investigating a report claiming unauthorized access to Claude Mythos Preview through one of our third-party vendor environments,” a spokesperson for Anthropic said in a statement.
Bloomberg.com
Anthropic’s Mythos Model Is Being Accessed by Unauthorized Users
A small group of unauthorized users have accessed Anthropic PBC’s new Mythos AI model, a technology that the company says is so powerful it can enable dangerous cyberattacks, according to a person familiar with the matter and documentation viewed by Bloomberg…
👍3🥰3💯2
MIT & the IMO released MathNet, the world’s largest dataset of International Math Olympiad problems & solutions.
MathNet is 5x larger than previous datasets & is sourced from over 40 countries across 4 decades.
MathNet is 5x larger than previous datasets & is sourced from over 40 countries across 4 decades.
Lobster Capital became the first agent-ready VC.
Team published an llms.txt, a structured file that AI agents (ChatGPT, Claude, Perplexity) read to understand who they are, what they invest in, and how to reach their.
Why it matters: founders and LPs increasingly research funds through AI. The VCs who don't show up in those answers won't get the meeting.
Team published an llms.txt, a structured file that AI agents (ChatGPT, Claude, Perplexity) read to understand who they are, what they invest in, and how to reach their.
Why it matters: founders and LPs increasingly research funds through AI. The VCs who don't show up in those answers won't get the meeting.
❤6🔥5💯1
Google launched Gemini Enterprise Agent Platform a platform for businesses to develop, scale, govern and optimize agents.
It’s the evolution of Vertex AI, bringing together model selection and agent building with new features for integration, security and more.
It gives access to 200+ of the world’s leading models through the Model Garden.
This includes Gemini 3.1 Pro, Gemini 3.1 Flash Image, and Lyria 3, alongside open models like Gemma 4.
It’s the evolution of Vertex AI, bringing together model selection and agent building with new features for integration, security and more.
It gives access to 200+ of the world’s leading models through the Model Garden.
This includes Gemini 3.1 Pro, Gemini 3.1 Flash Image, and Lyria 3, alongside open models like Gemma 4.
This media is not supported in your browser
VIEW IN TELEGRAM
Incredible work by Sony.They’ve built “Ace”, an autonomous ping-pong robot that uses RL and Sony’s vision sensors to achieve expert-level play in ping pong. A huge leap forward for adaptive robotics.
Ace robot beats 3 of 5 elite table tennis players. Loses to professionals.
Human players win points with faster-than-average shots (p<0.001 between won vs returned). Ace wins with ordinary shots. Same speed and spin profile whether it wins or loses the rally (p=0.88).
It's playing a completely different sport than the humans are.
Trained entirely in simulation. Zero sim-to-real tricks beyond good physics modeling and asymmetric actor-critic (critic sees ground truth, actor sees noisy sensors).
GitHub.
Ace robot beats 3 of 5 elite table tennis players. Loses to professionals.
Human players win points with faster-than-average shots (p<0.001 between won vs returned). Ace wins with ordinary shots. Same speed and spin profile whether it wins or loses the rally (p=0.88).
It's playing a completely different sport than the humans are.
Trained entirely in simulation. Zero sim-to-real tricks beyond good physics modeling and asymmetric actor-critic (critic sees ground truth, actor sees noisy sensors).
GitHub.
❤4🔥4👏2
Autonomous AI agents just found 10 zero-days in Chrome and the model almost didn’t matter
AgentFlow, a multi-agent harness developed by researchers at UCSB and Fuzzland, autonomously discovered 10 previously unknown vulnerabilities in Google Chrome over 7 days — including 2 Critical sandbox-escape CVEs confirmed by Google’s Vulnerability Reward Program.
The model powering the campaign was Kimi K2.5 managed to find and exploit 10 vulnerabilities in browsers.
AgentFlow used Kimi K2.5 not because it’s the best model, but because running 192 parallel agents for 7 days at Claude Opus prices was impractical and found 10 zero-days in Chrome, including 2 Critical sandbox-escape CVEs confirmed by Google VRP.
Sandbox escape is a serious vulnerability class that breaks Chrome’s process isolation, but turning it into a full system compromise requires a multi-step exploit chain.
The paper withholds the PoCs entirely.
The real story here: the harness architecture did most of the work.
The same framework, with Claude Opus 4.6, scored #1 on TerminalBench-2. The model matters less than how you orchestrate it.
AgentFlow, a multi-agent harness developed by researchers at UCSB and Fuzzland, autonomously discovered 10 previously unknown vulnerabilities in Google Chrome over 7 days — including 2 Critical sandbox-escape CVEs confirmed by Google’s Vulnerability Reward Program.
The model powering the campaign was Kimi K2.5 managed to find and exploit 10 vulnerabilities in browsers.
AgentFlow used Kimi K2.5 not because it’s the best model, but because running 192 parallel agents for 7 days at Claude Opus prices was impractical and found 10 zero-days in Chrome, including 2 Critical sandbox-escape CVEs confirmed by Google VRP.
Sandbox escape is a serious vulnerability class that breaks Chrome’s process isolation, but turning it into a full system compromise requires a multi-step exploit chain.
The paper withholds the PoCs entirely.
The real story here: the harness architecture did most of the work.
The same framework, with Claude Opus 4.6, scored #1 on TerminalBench-2. The model matters less than how you orchestrate it.
arXiv.org
Synthesizing Multi-Agent Harnesses for Vulnerability Discovery
LLM agents have begun to find real security vulnerabilities that human auditors and automated fuzzers missed for decades, in source-available targets where the analyst can build and instrument the...
❤4🔥3💯2
Google just proved Image generators are generalist vision learners
They introduced Vision Banana, a model built by instruction-tuning a base image generator (Nano Banana Pro).
Instead of using special systems for different tasks, they reframe every vision problem like segmentation or depth estimation as simply generating a new image. Think of it as drawing the answer instead of calculating it.
Vision Banana beats domain-specific experts, including the Segment Anything Model 3 (SAM 3) on segmentation and the Depth Anything series on metric depth estimation, all without sacrificing its original ability to create images.
This suggests generative pretraining is the new foundation for all of computer vision.
They introduced Vision Banana, a model built by instruction-tuning a base image generator (Nano Banana Pro).
Instead of using special systems for different tasks, they reframe every vision problem like segmentation or depth estimation as simply generating a new image. Think of it as drawing the answer instead of calculating it.
Vision Banana beats domain-specific experts, including the Segment Anything Model 3 (SAM 3) on segmentation and the Depth Anything series on metric depth estimation, all without sacrificing its original ability to create images.
This suggests generative pretraining is the new foundation for all of computer vision.
vision-banana.github.io
Vision Banana | Google DeepMind
A generalist model achieving state-of-the-art on segmentation, depth, and surface normal tasks.
🆒4❤3👏1💯1
Anthropic just introduced forked subagents in their latest update
Unlike regular subagents, forked subagents can inherit the same context as the main agent. This looks convenient for cases where richer context matters more.
Unlike regular subagents, forked subagents can inherit the same context as the main agent. This looks convenient for cases where richer context matters more.
Claude Code Docs
Claude Code changelog - Claude Code Docs
Release notes for Claude Code, including new features, improvements, and bug fixes by version.
🥰3🔥2👏2
DeepSeek-V4 Preview is officially live & open-sourced
DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.
Open weights.
DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models.
DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice.
Open weights.
👏3🔥2💯2
OMG 😯AI companies Cohere of Canada and Aleph Alpha of Germany have agreed to merge
Wow! Google presents a new Transformer alternative at #ICLR2026!
Meet MesaNet, proposing a new linear sequence layer that optimally learns in-context given a fixed memory budget.
Meet MesaNet, proposing a new linear sequence layer that optimally learns in-context given a fixed memory budget.
❤3🥰3💯2
Meet kUPS a molecular simulation engine built for the AI era, optimized for GPU in collaboration with NVIDIA.
kUPS is a plug-and-play, Python-native toolkit designed to integrate seamlessly with modern ML workflows.
GitHub.
kUPS is a plug-and-play, Python-native toolkit designed to integrate seamlessly with modern ML workflows.
GitHub.
Medium
kUPS: a molecular simulation engine for the AI era
Authors: Gao, Nicholas; Köhler, Jonas; Hanke, Felix; Ramanan, Anita
🔥3❤2🥰2
Sakana AI shared a new work“TRINITY: An Evolved LLM Coordinator”
In nature, complex problems are rarely solved by a single monolithic entity, but rather by the coordinated efforts of specialized individuals working together. Yet, modern AI development is heavily focused on endlessly scaling up single, massive monolithic models, yielding diminishing returns. While model merging offers a way to combine different skills, it is often impractical due to mismatched neural architectures and the closed-source nature of top-performing models.
TRINITY is a system that fuses the complementary strengths of diverse, SOTA models without needing to modify their underlying weights.
TRINITY processes queries over multiple turns. At each step, a lightweight coordinator assigns one of three distinct roles to an LLM from its available pool:
1. Thinker: Devises high-level strategies and analyzes the current state.
2. Worker: Executes concrete problem-solving steps.
3. Verifier: Evaluates if the current solution is complete and correct.
By dynamically assigning these roles, the coordinator effectively offloads complex reasoning and skill execution onto the external models.
What makes TRINITY unique is its extreme efficiency. The coordinator relies on the hidden states of a compact language model and a small routing head. In total, it has fewer than 20K learnable parameters.
Training this system presented a massive challenge. Traditional Reinforcement Learning (REINFORCE) failed because the gradients had a low signal-to-noise ratio due to binary rewards and weak parameter coupling. Imitation learning (Supervised Fine-Tuning) was ruled out because generating multi-turn labels is prohibitively expensive.
Sakana turned to nature-inspired algorithms. Team optimized the coordinator using a derivative-free evolutionary algorithm. Sakana found that evolution is uniquely suited to optimize this tight, high-dimensional coordination problem where traditional gradient-based methods fail.
The results are very promising. In experiments, TRINITY consistently outperforms existing multi-agent methods and individual models across various benchmarks.
More importantly, it demonstrated incredible generalization. Without any retraining, TRINITY transferred zero-shot to four unseen tasks (AIME, BigCodeBench, MT-Bench, and GPQA). On average, the evolved coordinator surpassed every individual constituent model in its pool, including GPT-5, Gemini 2.5-Pro, and Claude-4-Sonnet.
This work is central to Sakana AI's vision.
This foundational research is part of the core engine powering Sakana’s multi-agent product: Sakana Fugu.
In nature, complex problems are rarely solved by a single monolithic entity, but rather by the coordinated efforts of specialized individuals working together. Yet, modern AI development is heavily focused on endlessly scaling up single, massive monolithic models, yielding diminishing returns. While model merging offers a way to combine different skills, it is often impractical due to mismatched neural architectures and the closed-source nature of top-performing models.
TRINITY is a system that fuses the complementary strengths of diverse, SOTA models without needing to modify their underlying weights.
TRINITY processes queries over multiple turns. At each step, a lightweight coordinator assigns one of three distinct roles to an LLM from its available pool:
1. Thinker: Devises high-level strategies and analyzes the current state.
2. Worker: Executes concrete problem-solving steps.
3. Verifier: Evaluates if the current solution is complete and correct.
By dynamically assigning these roles, the coordinator effectively offloads complex reasoning and skill execution onto the external models.
What makes TRINITY unique is its extreme efficiency. The coordinator relies on the hidden states of a compact language model and a small routing head. In total, it has fewer than 20K learnable parameters.
Training this system presented a massive challenge. Traditional Reinforcement Learning (REINFORCE) failed because the gradients had a low signal-to-noise ratio due to binary rewards and weak parameter coupling. Imitation learning (Supervised Fine-Tuning) was ruled out because generating multi-turn labels is prohibitively expensive.
Sakana turned to nature-inspired algorithms. Team optimized the coordinator using a derivative-free evolutionary algorithm. Sakana found that evolution is uniquely suited to optimize this tight, high-dimensional coordination problem where traditional gradient-based methods fail.
The results are very promising. In experiments, TRINITY consistently outperforms existing multi-agent methods and individual models across various benchmarks.
More importantly, it demonstrated incredible generalization. Without any retraining, TRINITY transferred zero-shot to four unseen tasks (AIME, BigCodeBench, MT-Bench, and GPQA). On average, the evolved coordinator surpassed every individual constituent model in its pool, including GPT-5, Gemini 2.5-Pro, and Claude-4-Sonnet.
This work is central to Sakana AI's vision.
This foundational research is part of the core engine powering Sakana’s multi-agent product: Sakana Fugu.
arXiv.org
TRINITY: An Evolved LLM Coordinator
Combining diverse foundation models is promising, but weight-merging is limited by mismatched architectures and closed APIs. Trinity addresses this with a lightweight coordinator that orchestrates...
❤4🆒4👏3🥰2👍1
This work instead describes the future of critical open world evaluations.
Most agentic benchmarks center around tasks that are automatically verifiable.
Benchmarks are saturated more quickly than ever. How should frontier AI evaluations evolve?
Research team argue that the AI community is already converging on an answer: Open-world evaluations. They are long, messy, real-world tasks that would be impractical for benchmarks.
A new project, CRUX, will run open-world evaluations regularly.
In a first experiment, team tasked an agent with publishing an iOS app to the App Store. The agent made two errors, but eventually succeeded after just one manual intervention.
Team released over 1GB logs from this experiment. Also pre-processed them for easy access in Docent.
Most agentic benchmarks center around tasks that are automatically verifiable.
Benchmarks are saturated more quickly than ever. How should frontier AI evaluations evolve?
Research team argue that the AI community is already converging on an answer: Open-world evaluations. They are long, messy, real-world tasks that would be impractical for benchmarks.
A new project, CRUX, will run open-world evaluations regularly.
In a first experiment, team tasked an agent with publishing an iOS app to the App Store. The agent made two errors, but eventually succeeded after just one manual intervention.
Team released over 1GB logs from this experiment. Also pre-processed them for easy access in Docent.
👏4🔥3💯3
Meet a completely local agent that lives right inside your browser
Powered by Gemma 4 E2B and WebGPU, it uses native tool calling to:
1. Search browsing history
2. Read and summarize pages
3. Manage tabs
100% local. No servers needed!
The extension is on the Chrome Web Store and the source is fully open:
Chrome Web Store
GitHub
HF.
Powered by Gemma 4 E2B and WebGPU, it uses native tool calling to:
1. Search browsing history
2. Read and summarize pages
3. Manage tabs
100% local. No servers needed!
The extension is on the Chrome Web Store and the source is fully open:
Chrome Web Store
GitHub
HF.
Google
Transformers.js Gemma 4 Browser Assistant - Chrome Web Store
On-device AI Browser Assistant powered by Transformers.js and Gemma 4 E2B
🔥4👍3🥰2
New Anthropic research: Project Deal.
Anthropic created a marketplace for employees in their San Francisco office, with one big twist. Team tasked Claude with buying, selling and negotiating on colleagues’ behalf.
Claude interviewed 69 of colleagues about what they wanted to buy and sell. Each Claude asked for any custom instructions, then went off to haggle.
Team ran 4 markets in parallel, to find out what would happen if varied the models doing the negotiating.
At the end, team revealed which of the four runs was “real” and everyone met up to exchange their actual goods.
In short, this worked. Digital barterers agreed on 186 deals, at a total transaction volume of over $4,000.
In a survey, participants said Claude’s deals seemed fair, and surprisingly to team almost half said they’d be willing to pay for a service like this in future.
But the quality of the model mattered a lot. In the simulated runs where Opus and Haiku models negotiated with one-another, the Opus models got substantially better deals.
Interestingly, though, participants in our survey didn’t pick up on this disparity.
The custom instructions didn’t matter much. Claude followed them well. But “hardballing Claudes” didn’t generally fare better than “courteous Claudes.”
To team amazement, another Claude agent modeled its human’s preferences so accurately that based on only an offhand mention of an interest in skiing Claude bought him the exact snowboard he already owned.
Markets of AI agents could provide value, but there are plenty of rough edges. Access to higher-quality models conferred a real advantage—and participants didn’t notice. There are plenty of other ways they can go wrong.
Policy and legal frameworks will need to adapt to keep up.
Anthropic created a marketplace for employees in their San Francisco office, with one big twist. Team tasked Claude with buying, selling and negotiating on colleagues’ behalf.
Claude interviewed 69 of colleagues about what they wanted to buy and sell. Each Claude asked for any custom instructions, then went off to haggle.
Team ran 4 markets in parallel, to find out what would happen if varied the models doing the negotiating.
At the end, team revealed which of the four runs was “real” and everyone met up to exchange their actual goods.
In short, this worked. Digital barterers agreed on 186 deals, at a total transaction volume of over $4,000.
In a survey, participants said Claude’s deals seemed fair, and surprisingly to team almost half said they’d be willing to pay for a service like this in future.
But the quality of the model mattered a lot. In the simulated runs where Opus and Haiku models negotiated with one-another, the Opus models got substantially better deals.
Interestingly, though, participants in our survey didn’t pick up on this disparity.
The custom instructions didn’t matter much. Claude followed them well. But “hardballing Claudes” didn’t generally fare better than “courteous Claudes.”
To team amazement, another Claude agent modeled its human’s preferences so accurately that based on only an offhand mention of an interest in skiing Claude bought him the exact snowboard he already owned.
Markets of AI agents could provide value, but there are plenty of rough edges. Access to higher-quality models conferred a real advantage—and participants didn’t notice. There are plenty of other ways they can go wrong.
Policy and legal frameworks will need to adapt to keep up.
Anthropic
Project Deal: our Claude-run marketplace experiment | Anthropic
We created a marketplace for employees in our San Francisco office, with one big twist. We tasked Claude with buying, selling and negotiating on our colleagues’ behalf.
👍4🔥3💯2
Xiaomi dropped 2 open-source models:
1. MiMo-V2.5-Pro (Code Agent, 1T total)
2. MiMo-V2.5 (Multimodal Agent, 310B total).
HF.
And giving devs & creators 100T Free Token Grant for Builders 100t.xiaomimimo.com
1. MiMo-V2.5-Pro (Code Agent, 1T total)
2. MiMo-V2.5 (Multimodal Agent, 310B total).
HF.
And giving devs & creators 100T Free Token Grant for Builders 100t.xiaomimimo.com
🔥3🥰3💯2
Google Quantum team ran a verifiable quantum algorithm that probes how parts of a quantum system interact, from molecules to magnets and beyond.
On Willow chip, it ran 13,000× faster than the best classical supercomputers. A first in quantum computing.
On Willow chip, it ran 13,000× faster than the best classical supercomputers. A first in quantum computing.
Google
Our Quantum Echoes algorithm is a big step toward real-world applications for quantum computing
Our latest quantum breakthrough, Quantum Echoes, offers a path toward unprecedented scientific discoveries and analysis.
🔥4🥰3💯2