Page 22 / 25
291 posts in total. Keep on posting.
Showing posts 253–264 of 291. Each entry opens locally on this site; legacy Hexo posts link back to their original article at the bottom for reference.
2026
- EN
Self-Refine: Teaching LLMs to Critique and Improve Their Own Output — No Extra Training Needed
A technical review of Self-Refine, analyzing how iterative self-feedback loops enable LLMs to progressively improve their own outputs without external training or human supervision.
- EN
DeepSeekMath: How 120B Tokens of Math Data and GRPO Rival GPT-4 on Competition Problems
A detailed technical review of DeepSeekMath, analyzing how continued pretraining on math corpora combined with Group Relative Policy Optimization (GRPO) enables a 7B open model to rival frontier systems on challenging math benchmarks.
- EN
Reflexion: LLM Agents That Learn from Failure Through Verbal Self-Reflection
A technical review of Reflexion, exploring how language agents use verbal self-reflection as reinforcement signals to iteratively improve performance on coding, reasoning, and decision-making tasks.
- EN
vLLM and PagedAttention: Efficient Memory Management for Large Language Model Serving — Technical Review
A detailed technical review of the vLLM paper, which introduces PagedAttention—a novel attention algorithm inspired by OS virtual memory paging—to eliminate KV cache memory waste and dramatically increase LLM serving throughput.
- EN
AdaLoRA: Adaptive Budget Allocation for Parameter-Efficient Fine-Tuning — Technical Review
A detailed technical review of the AdaLoRA paper, which proposes adaptive budget allocation for parameter-efficient fine-tuning by parameterizing weight updates via SVD and dynamically pruning singular values based on importance scoring.
- EN
GLM-5 Technical Review: From Vibe Coding to Agentic Engineering
A detailed technical review of GLM-5's agentic engineering approach, analyzing how Zhipu AI moved from 'vibe coding' to systematic agent-driven software development with real-world deployment insights and lessons learned.
- EN
DeepSeek-V2: Multi-head Latent Attention and DeepSeekMoE — Technical Review
A detailed technical review of the DeepSeek-V2 architecture, focusing on Multi-head Latent Attention (MLA) which achieves 93.3% KV cache reduction through low-rank key-value joint compression, and DeepSeekMoE which enables economical training through fine-grained expert segmentation and shared expert isolation.
- EN
Direct Preference Optimization: Your Language Model Is Secretly a Reward Model — Technical Review
A detailed technical review of Rafailov et al.'s paper 'Direct Preference Optimization', analyzing how DPO eliminates the need for reinforcement learning in language model alignment by deriving a closed-form mapping from reward functions to optimal policies, enabling a simple classification loss to replace the complex RLHF pipeline.
- EN
Tree of Thoughts: Deliberate Problem Solving with Large Language Models — Technical Review
A detailed technical review of the Tree of Thoughts (ToT) framework, which generalizes chain-of-thought prompting to enable deliberate, search-based problem solving with large language models using BFS and DFS over structured reasoning trees.
- EN
ReAct Technical Review: From Reasoning Ability to Executable Reasoning
A comprehensive technical review of ReAct (Reasoning + Acting), analyzing how interleaving chain-of-thought reasoning with tool-use actions enables LLM agents to tackle complex tasks like question answering and web navigation with improved accuracy and interpretability.
2023
- EN
ComputerArchitecture-Day1
Notes on computer architecture fundamentals — covering CPU design, instruction sets, pipelining, and memory hierarchy basics.
2022
- EN
Reinforcement Learning-Principle-Day12
Reinforcement learning study notes — hierarchical RL, options framework, and goal-conditioned policies.