Tag archive
Evaluate SFT vs Reinforcement Learning for LLM reasoning. Benchmark compute, memory, and accuracy across GSM8K and MATH with hands-on code examples.
한국 AI 기업 GeniGenAI와 VIDRAFT가 LLM의 자기 추론 모니터링을 가능하게 하는 메타인지 어댑터를 공동 개발·공개했습니다. 재학습 없이 기존 모델에 플러그인 방식으로 적용 가능한 AGI 핵심 구성 요소를 확인하세요.
Originally published on The Searchless Journal The Invisible Leap If you used large...
An interactive explorable explanation of Monte Carlo Tree Search for LLM reasoning. Watch reasoning paths branch, dead-end, and backtrack.
New research shows RL post-training only modifies 1–3% of token positions, always within the base model's existing top-5 candidates. Here's what it means.
How multi-agent debate improves LLM factuality by 8+ points on math benchmarks. Paper breakdown and 90-line Python PoC implementation.
THINC trains a 4B parameter model to reason entirely in code. It scored 78.1% on competition math, beating Qwen3-235B at 75.2%. Here's how the method works.
ReFlect wraps any LLM with deterministic error detection and recovery at inference time. Hands-on guide with code and benchmark results from arXiv:2605.05737.
Key Takeaways OpenAI’s o1 models, released in September 2024, use large-scale reinforcement...