#GRPO training
42 related articles
·3 min
Why SFT Can't Fix the Root Cause of JSON Errors: How GRPO Correctness Training Breaks Through Coding Agent Bottlenecks
Analysis of why SFT can't fix coding agent JSON errors and how GRPO's binary reward signals and synchronized weight updates train directly for correctness.
Read more →
·4 min
Kimi K2/K3 Technical Speculation: Deep Dive into RL Hyperparameters and Multi-Teacher Distillation
Deep analysis of RL hyperparameter tuning challenges and 9-policy multi-teacher distillation in Kimi K2/K3 training, exploring the shift from scale to training craft.
Read more →
·3 min
A 9B Model Beats Frontier LLMs After Just $500 in RL Fine-Tuning
Just $500 in RL fine-tuning enables a 9B open-source model to outperform frontier LLMs on catalog review tasks. Analysis of when small-model RL works and its enterprise implications.
Read more →
·7 min
DeepSeek's Mysteriously Retracted Paper: Breaking AI's Multimodal Reasoning Bottleneck with Visual Primitives
DeepSeek's paper 'Thinking with Visual Primitives' was online for just 4 hours before being pulled. It uses bounding boxes and points as reasoning primitives, letting models 'point at' images to outperform GPT, Gemini, and Claude on maze navigation and counting.
Read more →
·3 min
Grok 4.5 Review: A Full Breakdown of Elon Musk's Most Powerful AI Coding Model
xAI releases Grok 4.5, ranking #1 on SWE Marathon and outperforming Claude Opus. Explore benchmark scores, Agent capabilities, free access, and CLI installation.
Read more →
·3 min
Beware of 'Free GPT' Scams: How to Spot AI Traffic-Baiting Schemes and Their Risks
Videos on Bilibili claiming free access to GPT-5.6 and Claude are traffic scams. Learn how to spot fake AI offers, understand the real risks, and find safe alternatives.
Read more →
·4 min
UK Sovereign AI: How Cosine AI Is Punching Above Its Weight Against American Compute Dominance
Cosine AI founder reveals how the UK's first sovereign LLM is being built — from government compute grants and RL credit attribution to multi-agent orchestration and synthetic data pipelines.
Read more →
·3 min
Qwen's Six-Generation Evolution: A Complete Breakdown from Architecture Innovations to the Reasoning Race
From Qwen1 to Qwen3-2507: a complete breakdown of six generations of evolution over two years — GQA, MoE, GRPO, dynamic thinking, and the core shift from architecture to reasoning.
Read more →
·2 min
Zhipu GLM New Model Coming Soon: Decoding the Founder's Teaser
Zhipu AI's Z.ai founder teases a new GLM model just one month after GLM 5.2. We decode the announcement, explore likely upgrade directions, and analyze China's LLM iteration pace.
Read more →
·2 min
Ollama Modelfile in Practice: Deploy Your Own Custom DeepSeek-R1 Model Locally
Step-by-step guide to wrapping DeepSeek-R1 with an Ollama Modelfile — set temperature, system prompts, and run fully offline for privacy and flexibility.
Read more →
·3 min
AI Large Language Models for Beginners: Core Concepts and Chinese Model Landscape
New to AI? This guide clarifies AI, machine learning, deep learning, and LLMs, traces milestones from Deep Blue to DeepSeek, and maps out China's LLM landscape.
Read more →
Three Core Gaps in Multimodal LLMs: Fr…
·4 min
Three Core Gaps in Multimodal LLMs: Frontier Research from Microsoft Research India
Microsoft Research India reveals three core gaps in multimodal LLMs: visual perception blindspots, cognitive hallucination, and architectural limitations. Explores Faithful GRPO, behavior modeling, and model alignment breakthroughs.
Read more →
·2 min
The Reasoning Trap in Diffusion Language Models: A One-Line Fix for the 'Flexibility' Flaw
ICML Outstanding Paper 'The Flexibility Trap' reveals how dLLMs' arbitrary-order generation bypasses logical connectives, breaking reasoning. JustGRPO fixes this with autoregressive training + parallel inference, achieving 89.1% on GSM8K in under 60 lines of code.
Read more →
·3 min
Can Open Source Topple ChatGPT? CSDN Founder Jiang Tao Breaks Down DeepSeek's Game-Changing Strategy
CSDN founder Jiang Tao analyzes how DeepSeek uses open source to challenge ChatGPT's dominance — from technical transparency to national AI strategy.
Read more →
·3 min
Meta Muse Spark 1.1 Deep Dive: The Platform Ambitions Behind Meta's First Closed-Source Model
Meta Muse Spark 1.1 deep dive: native multimodal architecture, platform tools, social data retrieval, e-commerce vision — Meta's first closed-source API model benchmarks against Anthropic Sonnet.
Read more →
·4 min
RAM: First-Principles Approach to 50x Faster RL Post-Training for Diffusion Models
RAM (Reinforce Adjoint Matching) achieves 50x faster RL post-training for diffusion models by discarding path costs, combining ODE sampling with decorrelated training objectives. A deep dive into RAM's core principles and experiments vs. Flow-GRPO.
Read more →
·2 min
iFLYTEK T30 Lite Learning Tablet In-Depth Review: Dual AI Engines + 12.5-Inch Eye-Care Display
The iFLYTEK T30 Lite learning tablet features the Spark large model and DeepSeek dual AI engines, a 12.5-inch eye-care display, and 8GB+512GB storage. This in-depth review analyzes its hardware, AI learning capabilities, and buying advice for K12 families.
Read more →
·3 min
Tencent Hunyuan HY3 In-Depth Review: How a Small Open-Source Model Punches Above Its Weight
In-depth analysis of Tencent's open-source reasoning model Hunyuan HY3: MoE architecture, 295B total params, Apache 2.0 license, coding & front-end rivaling DeepSeek V4 Pro at 1/35 the cost.
Read more →
·3 min
The Four-Stage Learning Path for AI Agent Development: A Complete Guide from Beginner to Practitioner
A systematic guide to the four-stage AI Agent development path: core concepts, principle paradigms like ReAct, RL and multi-agent optimization, and real-world projects. Mastering Agent development is the true hardcore edge in today's LLM field.
Read more →
·7 min
50 AI Research Job Applications to FAANG, 0 Replies? Unpacking the Truth Behind the Job Search Struggle
An AI research engineer with 3 years of experience sent 50 applications to FAANG with zero replies. This article breaks down the hidden barriers of top-tech AI roles, the truth about LinkedIn ghost jobs, and the MLE vs. Research Engineer divide.
Read more →