LuCiD-papers

Making AI papers lucid -- Learn, Code, Document

LCD Framework

Learn

Read and deeply understand the paper, its context, and contributions.

Code

Write visualization code hands-on -- practice by building static figures, interactive demos, and animations.

Document

Publish and share the work -- GitHub Pages demos, structured notes, and paper walkthroughs.

Papers

1706.03741 Done Alignment

Deep Reinforcement Learning from Human Preferences

Christiano, Leike, Brown, Martic, Legg, Amodei (2017) -- The foundational RLHF paper. Train RL agents using human preference comparisons instead of hand-designed reward functions.

5 PNGs 4 Interactive HTMLs 4 Animations

View on GitHub

Upcoming

Alignment Track

1707.06347 Done Alignment

Proximal Policy Optimization Algorithms

Schulman, Wolski, Dhariwal, Radford, Klimov (2017) -- PPO -- a simple, stable policy gradient method using a clipped surrogate objective. The RL optimizer used inside RLHF.

5 PNGs 6 Interactive HTMLs 5 Animations

View on GitHub

2009.01325 Coming Soon Alignment

Learning to Summarize from Human Feedback

Stiennon, Ouyang, Wu, Ziegler, Lowe, Voss, Radford, Amodei, Christiano (2020) -- Applies RLHF to text summarization, demonstrating preference-based training scales to NLP tasks.

2203.02155 Coming Soon Alignment

Training Language Models to Follow Instructions (InstructGPT)

Ouyang, Wu, Jiang, Almeida, Wainwright, Mishkin, Zhang, et al. (2022) -- RLHF at scale -- fine-tunes GPT-3 with human feedback to follow instructions.

2305.18290 Coming Soon Alignment

Direct Preference Optimization (DPO)

Rafailov, Sharma, Mitchell, Ermon, Manning, Finn (2023) -- Eliminates the reward model entirely -- optimizes preferences directly via a classification loss.

2402.03300 Coming Soon Alignment

Self-Play Fine-Tuning (SPIN)

Chen, Shen, Hong, Chen, Jiao, Zhang, Ma, Liu (2024) -- Aligns LLMs via self-play without human preference data.

VLM Track

2405.17247 Learning VLM

An Introduction to Vision-Language Modeling

Bordes, Pang, Ajay, et al. (2024) -- Comprehensive survey of VLM families: contrastive, masking, and generative approaches.