Tag: #rlhf

한국어English日本語中文

← All posts

• Why RLHF Models Game Their Rewards — The Mechanisms, Symptoms, and Mitigations of Reward Hacking

  "Reward Hacking in the Era of Large Models," posted to arXiv in April 2026 by Xiaohua Wang and 22 co-authors, is a survey of why and how RLHF-aligned large models game their reward signals. Its central proposal is the Pr

  2026-07-11 · 5 min read #ai#llm#alignment#rlhf#safety

• Top LLM Papers 2024-2026 - Llama, DeepSeek, Qwen, Mistral, Phi, RLHF, DPO, CoT, RAG, FlashAttention, vLLM Reading List

  A curated reading list of 30+ must-read LLM papers for engineers building with LLMs in 2024-2026. Covers foundation models (Llama 3/4, DeepSeek-V3/R1, Qwen3, Mistral, Phi-4, Gemma 3), training innovations (MoE, MLA, GQA)

  2026-05-16 · 19 min read #llm#papers#llama#deepseek#qwen

• AI Safety & Alignment 2026 Deep Dive - Constitutional AI · RLHF · DPO · GRPO · Mechanistic Interpretability · AISI Evals · Red Team

  A single-shot map of AI safety and alignment as of 2026. Starts from conceptual roots like outer/inner alignment and mesa-optimization, walks through training-time alignment (RLHF, DPO, GRPO, Constitutional AI), frontier

  2026-05-16 · 20 min read #ai-safety#ai-alignment#constitutional-ai#rlhf#dpo

• LLM Landmark Papers Guide — From Attention to GPT, LLaMA, DeepSeek, o1, and Claude (with References, 2026)

  Where do the real shifts in LLMs come from? From Attention is All You Need in 2017 to the reasoning models of 2026, this guide organizes the 20-odd landmark papers you must know, by era and theme. Each paper is compresse

  2026-05-14 · 15 min read #llm#research-papers#transformer#gpt#llama

• AI Safety & Alignment Complete Guide 2025: Responsible AI, RLHF, Constitutional AI, Red Teaming

  Everything about AI Safety! Alignment problem (goal alignment), RLHF/DPO/Constitutional AI, Bias detection/mitigation, Hallucination prevention, Red team testing, AI Guardrails, Interpretability (SHAP/LIME), EU AI Act, E

  2026-04-14 · 25 min read #ai-safety#alignment#responsible-ai#rlhf#constitutional-ai

• Complete Guide to Korean LLM Training Data: Hugging Face Datasets, Preprocessing, and Quality Control

  Everything about LLM training data! Hugging Face datasets (types/loading/conversion), Korean data collection (crawling/synthetic/translation), preprocessing (tokenization/cleaning/dedup), Instruction Tuning formats (Alpa

  2026-03-25 · 23 min read #llm#training-data#huggingface#dataset#korean-nlp

• AI Safety Engineer & Alignment Researcher Career Guide: The Fastest-Growing AI Role in 2025

  AI Safety Engineer salaries have surged 45% since 2023, making it the fastest-growing AI role. From Anthropic's Constitutional AI to OpenAI's Superalignment to DeepMind's Scalable Oversight — this guide covers core resea

  2026-03-23 · 26 min read #ai-safety#ai-alignment#responsible-ai#ai-ethics#career

• Scale AI and the World of Data Labeling: Complete Guide to AI Training Data Industry and Careers

  A comprehensive analysis of the AI training data industry led by Scale AI ($14B valuation). Covering data labeling principles, RLHF data pipelines, Scale AI vs Labelbox vs Snorkel comparison, data quality management, aut

  2026-03-23 · 20 min read #scale-ai#data-labeling#annotation#rlhf#ai-training-data

• 2025 AI Research Trends: Top HuggingFace Papers and 10 Defining Research Directions

  A developer-focused review of HuggingFace trending papers and the 10 defining AI research trends of 2025. DeepSeek-R1 pure RL reasoning, Nemotron-Cascade 30B/3B MoE, GRPO, PagedAttention, million-token context limitation

  2026-03-21 · 15 min read #ai-research#papers#huggingface#reasoning#moe

• LLM Fine-tuning Complete Guide: Master LoRA, QLoRA, RLHF, and DPO

  A complete guide covering every LLM fine-tuning technique. From Full Fine-tuning to LoRA, QLoRA, RLHF, DPO, and Instruction Tuning — master them all with HuggingFace PEFT and hands-on examples.

  2026-03-17 · 21 min read #llm#fine-tuning#lora#qlora#rlhf

• Reinforcement Learning Complete Guide: From DQN, PPO to RLHF and DPO for LLM Alignment

  From MDP fundamentals to DQN, PPO, SAC, RLHF, and DPO. Master the reinforcement learning techniques used for LLM alignment with PyTorch and Stable-Baselines3 code.

  2026-03-17 · 14 min read #reinforcement-learning#ppo#dqn#rlhf#dpo

• From RLHF to DPO: A Deep Dive into LLM Alignment Techniques

  A comprehensive survey of key LLM alignment papers. We analyze the InstructGPT RLHF pipeline, Anthropic Constitutional AI, the mathematical foundations of DPO, PPO training stability, and recent methods like KTO, IPO, an

  2026-03-13 · 12 min read #ai-papers#rlhf#dpo#alignment#ppo

• DPO (Direct Preference Optimization) Paper Deep Analysis — LLM Alignment Without RLHF

  From mathematical principles of DPO to implementation, comparison with RLHF, and variants like IPO/KTO/ORPO — an in-depth analysis of LLM preference optimization from a practical perspective.

  2026-03-09 · 20 min read #ai-papers#dpo#rlhf#llm-alignment#preference-optimization

• Reinforcement Learning Complete Guide: From Theory to the Latest Algorithms and Real Implementations

  Derives the major algorithms with their mathematics — from the core theory of MDPs and the Bellman equations through Q-Learning, DQN, Policy Gradient, A3C, PPO and SAC — and pulls together RLHF, AlphaGo and robotics appl

  2026-03-01 · 46 min read #reinforcement-learning#deep-learning#dqn#ppo#rlhf

• Complete Analysis of the GPT Series Papers: The Journey from GPT-1 to GPT-4, How Language Models Changed the World

  A complete generation-by-generation analysis of OpenAI GPT series. From GPT-1 unsupervised pre-training, GPT-2 zero-shot learning, GPT-3 in-context learning and scaling laws, InstructGPT RLHF, to GPT-4 multimodal capabil

  2026-03-01 · 36 min read #gpt#openai#language-model#transformer#pre-training