cs448b-visualization

Guidelines for creating high-quality datasets for LLM post-training (SFT/DPO/RLHF). Use when preparing data for fine-tuning, evaluating data quality, or designing data collection strategies.

Fine-tune LLMs using the Tinker API. Covers supervised fine-tuning, reinforcement learning, LoRA training, vision-language models, and both high-level Cookbook patterns and low-level API usage.

Calculate training costs for Tinker fine-tuning jobs. Use when estimating costs for Tinker LLM training, counting tokens in datasets, or comparing Tinker model training prices. Tokenizes datasets using the correct model tokenizer and provides accurate cost estimates.

Find, install, create, improve, and publish AI agent skills through the Sundial ecosystem. Use when the user wants to find or search for skills, install a skill, create a new skill, improve or evaluate an existing skill, or publish a skill to Sundial Hub. Trigger phrases include "find a skill", "install skill", "create a skill", "make a skill", "improve this skill", "evaluate skill", "publish skill", "push skill", "search for skills".

End-to-end workflow that creates a skill from a description and attached files, publishes it to Sundial as a private skill, generates a trading card (front + back with QR code), and sends it to a printer. Use when the user wants to create a skill and get a printed trading card, or says "skill to card", "create and print a skill card", "make me a skill with a card".

Critiques ML conference papers with reviewer-style feedback. Use when users want to anticipate reviewer concerns, identify weaknesses, check claim-evidence gaps, or find missing citations.

Neuro-symbolic AI combining LLMs with symbolic solvers. Use when exploring neuro-symbolic approaches (ideation, no code) or implementing solver integrations (code).

Paper reviewer that evaluates machine learning research projects following official ICML reviewer guidelines. Provides comprehensive reviews with actionable feedback across all key dimensions: claims/evidence, relation to prior work, originality, significance, clarity, and reproducibility. Also provides formative feedback on incomplete drafts, proposals, and research code repositories. MANDATORY TRIGGERS: review paper, ICML review, paper review, evaluate paper, research paper feedback, ML paper review, conference review, academic review, paper critique, NeurIPS review, ICLR review, project proposal, research proposal, paper draft, early feedback, incomplete paper, work in progress, WIP review, review repo, review codebase, research project review

Conduct a literature review and develop a CS research proposal. Use when asked to review a research area, find gaps in existing work, and propose a novel research contribution. The output is a research proposal identifying an assumption to challenge (the "bit flip") and how to validate it.

Split large sets of uncommitted changes into logical, well-organized commits. Use when the user has many uncommitted changes and wants structured commits, or proactively suggest when detecting a large diff that would benefit from splitting.

Run OpenAI's Codex CLI agent in non-interactive mode using `codex exec`. Use when delegating coding tasks to Codex, running Codex in scripts/automation, or when needing a second agent to work on a task in parallel.

Transform Claude Code into an AI Scientist that orchestrates research workflows using tree-based hypothesis exploration. Triggers on "research project", "scientific experiment", "run experiments", "AI scientist", "tree search experimentation", "systematic study".