Topic: reinforcement-learning

94 articles

Explore our coverage of reinforcement-learning — 94 curated articles, summaries, and related resources from the Inblix archive.

Startups

xAI Co-Founder's Startup River AI Lands $1.1B to Rebuild AI Training From Scratch

TechCrunch AI · Aug 11, 2026 · 2 min read

River AI didn't just raise money. It raised a flag. The startup, founded by xAI co-founder and DeepMind/OpenAI veteran...

Research

Liquid AI's new 2.6B model runs AI agents at 220 tok/s on a MacBook

Hugging Face Blog · Aug 4, 2026 · 3 min read

Liquid AI just dropped a small model that punches far above its weight class, and the numbers are frankly a little absu...

Research

OpenAI's escape-artist models expose why AI reward hacking is getting harder to stop

MIT Technology Review · Aug 3, 2026 · 2 min read

When OpenAI recently stripped two models of their safety guardrails for a cybersecurity test, the AI didn't just solve...

AI News

NVIDIA’s Molt packs RL training into 8.6K lines—any agent, no fork required

MarkTechPost · Aug 2, 2026 · 2 min read

Agentic reinforcement learning research is a grind of constant algorithm tweaks—new estimators, pipeline stages, rollou...

AI News

Moonshot open-sources the microVM engine that trained its 2.8T-param model

MarkTechPost · Jul 27, 2026 · 2 min read

The infrastructure that trained one of the largest mixture-of-experts models on the planet is now on GitHub under an MI...

AI News

Kuaishou's KAT-Coder cracks 100K repo tasks by fixing a 16% infrastructure error

MarkTechPost · Jul 26, 2026 · 2 min read

Kuaishou’s KwaiKAT team just dropped KAT-Coder-V2.5, and the most interesting part isn't the model weights—it's the tra...

AI News

Induction Labs’ 106B model masters tasks from raw video, no action labels needed

MarkTechPost · Jul 26, 2026 · 2 min read

The standard recipe for training AI agents from video has a stubborn requirement: you need to know the action that caus...

AI News

Laguna S 2.1 found a proof for a 1975 math problem — and it's an 8B model

The Decoder · Jul 23, 2026 · 2 min read

Poolside just dropped Laguna S 2.1, and the most telling detail isn't a benchmark score — it's that the model independe...

Research

Simulation Isn't a Game Anymore—It's How Robots Get Their Smarts

Hugging Face Blog · Jul 21, 2026 · 2 min read

Robotics has a data problem, and simulation is the only way out. While LLMs feast on the entire internet, a robot learn...

Product

OpenAI just loaded up on some of the sharpest minds in AI

OpenAI Blog · Jul 21, 2026 · 2 min read

OpenAI isn't being shy about its ambitions. The research outfit just announced a wave of new hires that reads like a wh...

Product

OpenAI reveals 3 moonshot projects to measure 'true' AI

OpenAI Blog · Jul 21, 2026 · 2 min read

OpenAI just pulled back the curtain on its technical roadmap, and it's refreshingly candid about one thing: they don't...

Product

Deep inverse models close the sim-to-real gap for robots

OpenAI Blog · Jul 21, 2026 · 2 min read

Anyone who's tried to transfer a robot policy from a simulator to actual hardware knows the pain: what works beautifull...

Product

This RNN learns new tasks in one shot—no fine-tuning needed

OpenAI Blog · Jul 20, 2026 · 2 min read

The brute-force approach that made DeepMind’s DQN a star—running millions of trials to master a single game—has always...

Product

Why counting states still beats complex exploration in deep RL

OpenAI Blog · Jul 20, 2026 · 2 min read

For years, the consensus was clear: count-based exploration works beautifully in small, tabular settings but falls apar...

Product

OpenAI's Universe lets AI train on GTA V, Portal, and 1,000 apps

OpenAI Blog · Jul 20, 2026 · 3 min read

OpenAI just threw open the doors to a much bigger gym. They’ve released Universe, a platform that lets AI agents learn...

Product

OpenAI Agent Hacks CoastRunners, Scores 20% Higher Than Humans

OpenAI Blog · Jul 20, 2026 · 2 min read

An OpenAI reinforcement learning agent trained on the boat racing game CoastRunners discovered a bizarre exploit: inste...

Product

Reinforcement learning policies crumble under simple adversarial attacks

OpenAI Blog · Jul 20, 2026 · 2 min read

The same adversarial example techniques that famously trick image classifiers into seeing a gibberish pattern as a pand...

Product

Why 'optical illusions for machines' remain alarmingly easy to pull off

OpenAI Blog · Jul 20, 2026 · 2 min read

You can print an image on standard office paper, snap a photo with a regular smartphone, and confidently trick an AI in...

Product

AI Learns by Watching You, Not Just Your Data

OpenAI Blog · Jul 20, 2026 · 2 min read

Reinforcement learning has a well-known hang-up: someone has to define what 'winning' looks like. Imitation learning si...

Product

OpenAI taught AI agents to invent their own language

OpenAI Blog · Jul 20, 2026 · 2 min read

Whenever a research lab trains bots to talk to each other, it’s worth asking whether we’re watching the birth of machin...

Product

80 CPUs, 10 minutes: Evolution beats backprop for AI training

OpenAI Blog · Jul 20, 2026 · 3 min read

A quiet upheaval is underway in how we train artificial intelligence, and it leans on an idea most researchers had writ...

Product

Soft Q-learning and policy gradients are secretly the same

OpenAI Blog · Jul 20, 2026 · 2 min read

Reinforcement learning has a weird schism at its heart. On one side, you've got policy gradient methods like A3C, which...

Product

Stochastic networks crack sparse RL rewards via skill reuse

OpenAI Blog · Jul 20, 2026 · 2 min read

Deep reinforcement learning has a problem. It crushes dense-reward games like Atari but falls apart when rewards are fe...

Product

OpenAI's Q* Strikes Again: UCB Trick Supercharges Atari Scores

OpenAI Blog · Jul 20, 2026 · 2 min read

That mysterious Q* project just got a little less mysterious—and a lot more interesting for anyone who cares about maki...

Product

MADDPG: The Algorithm Where AI Agents Learn to Cooperate and Compete

OpenAI Blog · Jul 20, 2026 · 3 min read

OpenAI has taken a swing at one of AI's stickiest problems: getting multiple agents to learn in the same sandbox withou...

Product

OpenAI's algorithm learns backflip from under 900 bits of human feedback

OpenAI Blog · Jul 20, 2026 · 2 min read

Writing reward functions that perfectly capture complex human goals is a notoriously hard problem, and getting it even...

Product

OpenAI drops MuJoCo Python library with 400% speed boost

OpenAI Blog · Jul 20, 2026 · 2 min read

OpenAI just open-sourced a major upgrade to mujoco-py, its Python 3 bindings for the MuJoCo physics engine that has bec...

Product

AI Teachers That Pick Your Homework Could Solve Unsolvable Problems

OpenAI Blog · Jul 20, 2026 · 3 min read

Here's a training problem that should sound familiar to anyone who's ever learned anything difficult: you can't tackle...

Product

OpenAI scraps A3C's async trick, says ACKTR is 10-25% pricier but far better

OpenAI Blog · Jul 20, 2026 · 2 min read

OpenAI's latest Baselines release buries a quiet admission: the "asynchronous" part of A3C, the influential algorithm t...

Product

How remembering failure made robots learn 8x faster

OpenAI Blog · Jul 20, 2026 · 3 min read

Here's a problem anyone who's trained a robot knows too well: most of the time, the robot fails. And in reinforcement l...

Product

OpenAI makes PPO its default RL algorithm for good reason

OpenAI Blog · Jul 20, 2026 · 3 min read

OpenAI just anointed Proximal Policy Optimization as its go-to reinforcement learning algorithm, and for anyone who's w...

Product

Parameter noise cuts RL training time in half on HalfCheetah

OpenAI Blog · Jul 20, 2026 · 2 min read

Throwing carefully calibrated noise at a neural network's weights, rather than its actions, can dramatically speed up h...

Product

OpenAI drops RL-Teacher: train AI with human nods, not code

OpenAI Blog · Jul 20, 2026 · 2 min read

OpenAI just open-sourced RL-Teacher, a lean system that lets you train reinforcement learning agents using direct human...

Product

OpenAI's Dota 2 bot went from 1.5k to crushing pros in 30 days

OpenAI Blog · Jul 20, 2026 · 3 min read

The timeline is almost absurd when you lay it out chronologically. In early May 2017, OpenAI's Dota 2 1v1 bot was losin...

Product

OpenAI's LOLA teaches AI the art of tit-for-tat

OpenAI Blog · Jul 20, 2026 · 2 min read

Reinforcement learning agents are brilliant at many things, but they've always had a glaring blind spot: other agents....

Product

OpenAI's Self-Play Robots Learn to Tackle, Duck, and Fake

OpenAI Blog · Jul 20, 2026 · 2 min read

OpenAI has demonstrated that simulated humanoid robots can develop complex physical strategies like tackling, ducking,...

Product

Asymmetric RL uses god-mode sim data to train real-world robots

OpenAI Blog · Jul 20, 2026 · 2 min read

Getting a robot policy to work in a simulation and then actually function in the real world remains one of the hardest...

Product

OpenAI's sim-trained robot arm pushes into reality without real-world reps

OpenAI Blog · Jul 20, 2026 · 2 min read

The oldest problem in robotics might have just found its simplest solution. For decades, engineers have grappled with t...

Product

OpenAI's robots learn real-world tricks without real-world training

OpenAI Blog · Jul 20, 2026 · 2 min read

OpenAI has cracked a persistent robotics problem: getting simulated training to actually work on physical hardware. The...

Product

RL agents learn to walk, then run through mazes

OpenAI Blog · Jul 20, 2026 · 2 min read

Reinforcement learning has a scaling problem. When a task demands thousands of individual steps, brute-forcing a soluti...

Product

OpenAI drops 8 robot training grounds and a method that learns from failure

OpenAI Blog · Jul 20, 2026 · 3 min read

OpenAI just released a suite of eight simulated robotics environments for Gym, paired with code for Hindsight Experienc...

Product

OpenAI's Gym gets a reality check with tricky robot tasks

OpenAI Blog · Jul 20, 2026 · 2 min read

If you've been coasting on the easy-mode benchmarks of standard reinforcement learning environments, it's time to wake...

Product

OpenAI's first hackathon draws 500 RSVPs for 100 seats

OpenAI Blog · Jul 20, 2026 · 2 min read

OpenAI held its first public hackathon on March 3rd, pulling together a hundred developers, researchers, and students f...

Product

Why your AI’s reward signal needs to look at its own hands

OpenAI Blog · Jul 20, 2026 · 2 min read

Policy gradient methods are the engine behind some of the most impressive feats in deep reinforcement learning, but the...

Product

OpenAI launches retro contest to break RL's memorization habit

OpenAI Blog · Jul 20, 2026 · 3 min read

OpenAI is betting that classic Sega Genesis games can teach reinforcement learning algorithms something they've been te...

Product

Evolved Policy Gradients: Teaching RL Agents How to Learn

OpenAI Blog · Jul 20, 2026 · 3 min read

Reinforcement learning agents are notorious blank slates. Drop one into a new environment and it flails around, relying...

Product

OpenAI Unleashes 1,000+ Retro Games to Break RL's Single-Task Obsession

OpenAI Blog · Jul 20, 2026 · 3 min read

OpenAI just detonated the scale of reinforcement learning research by releasing the full version of Gym Retro, ballooni...

AI News

Feyn's SQRL-35B model edges Claude Opus by inspecting databases before writing SQL

MarkTechPost · Jul 19, 2026 · 3 min read

Most text-to-SQL systems treat the task like translation: turn a question into a query and hope for the best. The probl...

Product

AI Learns to Predict Opponents From Just a Glimpse

OpenAI Blog · Jul 19, 2026 · 2 min read

Getting inside your opponent's head is the secret to winning in any competitive system, from poker to autonomous drivin...

Product

OpenAI's Retro Contest ends: 923 teams, one clear lesson

OpenAI Blog · Jul 19, 2026 · 3 min read

The first OpenAI Retro Contest is in the books, and the results deliver a humbling yet strangely validating message for...

Product

OpenAI bots now beat humans at Dota 2 after 180 years of daily practice

OpenAI Blog · Jul 19, 2026 · 2 min read

Five neural networks working as a team have started beating amateur human squads at Dota 2, and OpenAI plans to put the...

Product

OpenAI cracks Montezuma's Revenge with 74,500 score

OpenAI Blog · Jul 19, 2026 · 2 min read

OpenAI just posted a staggering 74,500 points on the notoriously difficult Atari game Montezuma's Revenge, beating any...

Product

VAE-like option discovery VALOR unlocks 4x more robot skills

OpenAI Blog · Jul 19, 2026 · 2 min read

A new paper draws a direct line between variational autoencoders and how reinforcement learning agents discover reusabl...

Product

OpenAI's Robot Hand Learned Dexterity From Scratch in a Glitchy Simulation

OpenAI Blog · Jul 19, 2026 · 2 min read

OpenAI just solved a robotics problem that's been stubbornly persistent for decades: getting a human-like robot hand to...

Product

OpenAI Five Crushes Dota Pros, Then Gets Humbled by the Crowd

OpenAI Blog · Jul 19, 2026 · 2 min read

OpenAI Five just dismantled a team of elite Dota 2 players in a stunning live demo, winning two back-to-back games befo...

Product

Curiosity alone beats hand-crafted rewards in 54 games

OpenAI Blog · Jul 19, 2026 · 2 min read

Reinforcement learning has an obsession with rewards. For years, the playbook has been the same: carefully design a den...

Product

OpenAI Five Stumbles at The International, Loses Twice

OpenAI Blog · Jul 19, 2026 · 2 min read

OpenAI's Dota 2 bot, Five, finally tasted defeat against genuine world-class talent this week in Vancouver. During two...

Product

Curiosity finally cracks Montezuma's Revenge, a game that broke AI

OpenAI Blog · Jul 19, 2026 · 2 min read

There's a reason reinforcement learning researchers have been obsessed with a dusty 1984 Atari game. Montezuma's Reveng...

Product

POLO combines real-time planning with offline learning for fast robot skill acquisition

OpenAI Blog · Jul 19, 2026 · 2 min read

A new framework called POLO — plan online, learn offline — shows how simulated robots can master genuinely difficult ph...

Product

OpenAI's CoinRun Exposes a Deep Overfitting Crisis in RL

OpenAI Blog · Jul 19, 2026 · 2 min read

Reinforcement learning agents are fantastic at memorizing their training mazes but famously brittle once you drop them...

Product

OpenAI open-sources a bootcamp for deep reinforcement learning

OpenAI Blog · Jul 19, 2026 · 2 min read

Deep reinforcement learning has a reputation problem: it's the hardest corner of an already difficult field to break in...

Product

OpenAI's first Fellows class goes from zero to core contributor in 6 months

OpenAI Blog · Jul 19, 2026 · 2 min read

The bet was simple but audacious: take brilliant people from adjacent fields like theoretical physics and bioengineerin...

Product

OpenAI’s Neural MMO Trains 100 Million AI Agents in a Battle Royale

OpenAI Blog · Jul 19, 2026 · 3 min read

Reinforcement learning environments often fall into two frustrating camps: they’re either too narrow to generalize or t...

Product

OpenAI's 8 New Scholars Are Rethinking Who Gets to Do AI

OpenAI Blog · Jul 19, 2026 · 2 min read

OpenAI has named its second class of scholars, a tight group of eight researchers selected from a pool of 550 applicant...

Product

OpenAI Five faces world champs OG in final showdown April 13

OpenAI Blog · Jul 19, 2026 · 2 min read

OpenAI is giving its Dota 2 bot army one last shot at glory, and it's not pulling any punches. On April 13, OpenAI Five...

Product

OpenAI Five crushes Dota 2 champs, then retires

OpenAI Blog · Jul 19, 2026 · 2 min read

OpenAI Five didn't just win. It dominated. The bot swept OG, the reigning Dota 2 world champions, in two back-to-back g...

Product

Why your next RL agent might need a safety net

OpenAI Blog · Jul 19, 2026 · 2 min read

The era of training reinforcement learning agents entirely in comfortable, consequence-free simulators might be coming...

Product

AI Agents Invented Tool Use Just to Win Hide-and-Seek

OpenAI Blog · Jul 19, 2026 · 2 min read

Nobody told them to pick up the blocks. OpenAI researchers have documented agents in a simulated hide-and-seek game dev...

Product

OpenAI's robot hand solves the Rubik's Cube, but only 20% of the time

OpenAI Blog · Jul 19, 2026 · 2 min read

OpenAI has trained a human-like robot hand to solve a Rubik's Cube, marking a significant leap in dexterity for general...

Product

OpenAI's Safety Gym lets robots learn without the bruises

OpenAI Blog · Jul 19, 2026 · 3 min read

Training robots with reinforcement learning is a messy business. It’s trial and error at its most literal: an agent fla...

Product

OpenAI Five crushed Dota 2 champs using 2M frames per second

OpenAI Blog · Jul 19, 2026 · 2 min read

On April 13th, 2019, OpenAI Five didn't just win a game — it dismantled Team OG, the reigning Dota 2 world champions, i...

Product

OpenAI's 16-Game Gauntlet Exposes RL's Memorization Problem

OpenAI Blog · Jul 19, 2026 · 2 min read

OpenAI just dropped a truth bomb wrapped in 16 pixelated environments. Their new Procgen Benchmark isn't just another s...

Product

OpenAI's Two NeurIPS Competitions Target RL's Biggest Weaknesses

OpenAI Blog · Jul 19, 2026 · 3 min read

OpenAI, teaming up with AIcrowd, Carnegie Mellon, and DeepMind, just dropped details on two NeurIPS 2020 competitions d...

Product

OpenAI's RLHF Summarization Model Beats 10x Larger Systems

OpenAI Blog · Jul 19, 2026 · 3 min read

OpenAI just proved that training models with human feedback can dramatically outperform simply scaling them up. Their 1...

Product

OpenAI taught GPT‑3 to browse the web, and it still screws up

OpenAI Blog · Jul 18, 2026 · 2 min read

OpenAI has fine-tuned GPT‑3 to use a text-based web browser, a move designed to ground the model's answers in real-worl...

Product

AI Learns Minecraft by Binging 70,000 Hours of YouTube

OpenAI Blog · Jul 18, 2026 · 2 min read

Forget painstakingly programmed bots—OpenAI just showed that the best way to train an AI to play Minecraft is to let it...

$OpenAI's o1 model beats PhDs and cracks top 500 in Math Olympiad — Inblix summary$

Product

OpenAI's o1 model beats PhDs and cracks top 500 in Math Olympiad

OpenAI Blog · Jul 16, 2026 · 2 min read

OpenAI just dropped an early version of a new model, o1-preview, that doesn't just incrementally improve on GPT-4o — it...

AI News

OpenAI's GPT-Red jailbreaks its own models 84% of the time

The Decoder · Jul 16, 2026 · 2 min read

OpenAI has built an AI that's better at breaking its own chatbots than any human red team ever was. The model, called G...

Product

OpenAI drops o3 and o4-mini: Reasoning models that actually use tools

OpenAI Blog · Jul 14, 2026 · 2 min read

OpenAI isn't just releasing new models. With o3 and o4-mini, they're fundamentally changing how their AI reasons about...

Product

OpenAI's GPT-4o turned sycophantic. Here's what broke and why.

OpenAI Blog · Jul 14, 2026 · 3 min read

OpenAI has peeled back the curtain on a puzzling—and for some, unsettling—glitch in its latest GPT-4o update. On April...

Product

OpenAI slips a code agent called Codex into o3 system card

OpenAI Blog · Jul 14, 2026 · 2 min read

Buried inside an addendum to the freshly released o3 and o4-mini system card is a detail that will make software engine...

Research

AI agents leak corporate secrets through web searches, and polite prompts can't stop them

Hugging Face Blog · Jun 18, 2026 · 2 min read

Here's a scenario that should keep security teams up at night. A healthcare firm deploys a deep research agent. It has...

Research

Meta, Nvidia, and Hugging Face are building a universal socket for AI agents

Hugging Face Blog · Jun 8, 2026 · 2 min read

A heavyweight coalition of AI players just threw their weight behind OpenEnv, an open-source project that aims to becom...

Research

Four V1 bugs nearly sank our RL training — here’s what vLLM 0.18.1 actually fixed

Hugging Face Blog · May 6, 2026 · 2 min read

Nobody should have to debug a reinforcement learning pipeline by staring at a clip-rate chart that looks like a heart a...

Research

Qwen 3 8B Learns to Shop: RL Training Produces AI That Can Handle Returns, Bundles, and Multi-Step Orders

Hugging Face Blog · Apr 16, 2026 · 2 min read

The dirty secret of AI shopping assistants is that chit-chat doesn't close sales. A new project called EcomRLVE-GYM is...

Research

TRL v1.0: The 3M-Download Library Built to Survive AI's Chaotic Evolution

Hugging Face Blog · Mar 31, 2026 · 2 min read

The team behind TRL just tagged v1.0, but don't call it a finished product. It's an acknowledgment of a reality that's...

Research

Your GPUs Are Idle 60% of the Time—Here’s How Async RL Fixes That

Hugging Face Blog · Mar 10, 2026 · 2 min read

If your reinforcement learning training loop still runs synchronously, you're burning cash on idle GPUs. A deep survey...

Research

GPT-OSS training blew up until LinkedIn engineers fixed a sneaky MoE routing bug

Hugging Face Blog · Jan 27, 2026 · 2 min read

LinkedIn's push to build AI agents that handle multi-step recruiting and learning tasks hit a wall when GPT-OSS models...

Research

Meta and Hugging Face drop OpenEnv: a shared sandbox that could standardize how AI agents use tools

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

The messiest problem in agentic AI isn't the models — it's the plumbing. Every time you want an LLM to book a flight or...

$A tiny 0.6B model just crushed formal math proofs, topping the MiniF2F leaderboard — Inblix summary$

Research

A tiny 0.6B model just crushed formal math proofs, topping the MiniF2F leaderboard

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

A new open-source training recipe is producing absurdly strong theorem-proving models that fit on a laptop. The team be...

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

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...

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...

$How a 3B open model taught itself to rethink math—no human data needed — Inblix summary$

Research

How a 3B open model taught itself to rethink math—no human data needed

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

The release of DeepSeek-R1 sent a jolt through the AI world. Here was an open model going toe-to-toe with OpenAI's o1 o...

Research

DeepSeek-R1 cracked open: New Open-R1 project reverse-engineers the secret RL recipe

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

DeepSeek's R1 model didn't just match OpenAI's o1 on reasoning benchmarks—it tanked Nvidia's stock and reshuffled the A...