training

RSS

1
Learning to solve hard problems in RL for LLMs by never giving up github.io
The post introduces the *Matthew Effect* in RL‑fine‑tuning of LLMs—performance gains concentrate on tasks the model already solves— and proposes *Never Give Up* (NGU), an adaptive sampling scheme that uses a small k for easy prompts and retries hard prompts with a high‑probability “never give up” loop. Experiments on math (AIME, GSM8k), code (Manufactoria), and larger‑scale setups (DeepScaler) sh…
Hacker News front pagegithub.ioMichael Noukhovitch6d11 minHN1179
2
Landing the Space Shuttle – A Flying Machine and the Thrill of a Lifetime eaa.org
A former NASA astronaut recounts the dead-stick landing of the Space Shuttle Atlantis from Mach 1 and 50,000 feet, a high-stakes maneuver with no go-around option. Extensive training, including 1,700 simulator approaches in a modified Gulfstream G-II, was crucial for mastering the shuttle's unique flight characteristics.
Hacker News front pageeaa.orgEAA6d6 minHN409
3
DACA-GRPO: Denoising-Aware Credit Assignment for Reinforcement Learning in Diffusion Language Models apple.com
DACA‑GRPO adds denoising‑aware credit assignment to GRPO‑style RL trainers for diffusion LLMs. It computes per‑token importance scores from intermediate denoising steps and uses stratified masking to reduce mean‑field bias in likelihood estimates. Plug‑and‑play on three existing GRPO methods, it yields consistent gains on seven downstream tasks (up to +5.6 pp math, +7.4 pp code, +36.3 pp constrai…
Apple Machine Learning Researchapple.com6d1 minpaper
4
1 points
MiniCPM5-2B Ranks First Among Open-Weight Models Under 4B sitepoint.com
MiniCPM5-2B, a 2.6 B‑parameter dense Llama‑style model, tops the Artificial Analysis GDPval‑AA v2 benchmark (831 Elo) and sits on the Pareto frontier of the Intelligence Index v4.2 despite being far smaller than competing models. The post breaks down its capability density, token‑cost efficiency, architecture, and three‑stage training (SFT, RL, on‑policy distillation) to explain why it outperform…
SitePointsitepoint.comNaveed Wali Khan6d10 min
5
How to upskill enterprise AI builders by using daily micro habits google.com
Google Cloud Consulting proposes a four‑pillar micro‑learning framework for enterprise AI upskilling: 5‑minute browser‑based exercises, pre‑configured sandboxes, daily streaks, and delivering runnable code each session. A pilot (Advent of Agents) showed >150k participants, 859k code runs, and a 31% daily return rate, suggesting short, frictionless tasks improve engagement versus traditional bootc…
Google Cloud Bloggoogle.comRyan Faris4d3 min

training

Learning to solve hard problems in RL for LLMs by never giving upgithub.io

Landing the Space Shuttle – A Flying Machine and the Thrill of a Lifetimeeaa.org

DACA-GRPO: Denoising-Aware Credit Assignment for Reinforcement Learning in Diffusion Language Modelsapple.com

MiniCPM5-2B Ranks First Among Open-Weight Models Under 4Bsitepoint.com

How to upskill enterprise AI builders by using daily micro habitsgoogle.com

Learning to solve hard problems in RL for LLMs by never giving up github.io

Landing the Space Shuttle – A Flying Machine and the Thrill of a Lifetime eaa.org

DACA-GRPO: Denoising-Aware Credit Assignment for Reinforcement Learning in Diffusion Language Models apple.com

MiniCPM5-2B Ranks First Among Open-Weight Models Under 4B sitepoint.com

How to upskill enterprise AI builders by using daily micro habits google.com