THUDM

9 mods across 2 repositories, 400 stars between them.

add-dynamic-filter

01

THUDM/DeepDive

Skill Claude CodeCodex

Guide for adding dynamic/filter hooks in slime rollout pipeline. Use when user wants sample-group selection during rollout, buffer filtering before training, or per-sample masking/processing hooks.

345 2mo ago A 40 tokens

THUDM/DeepDive

Skill Claude CodeCodex

Guide for adding and validating evaluation dataset configuration in slime. Use when user wants to configure eval datasets via --eval-config or --eval-prompt-data, add per-dataset overrides, or customize evaluation rollout behavior.

345 2mo ago A 49 tokens

add-reward-function

03

THUDM/DeepDive

Skill Claude CodeCodex

Guide for adding a custom reward function in slime and wiring it through --custom-rm-path (and optional reward post-processing). Use when user wants new reward logic, remote/service reward integration, or task-specific reward shaping.

345 2mo ago A 49 tokens

THUDM/DeepDive

Skill Claude CodeCodex

Guide for adding a new rollout function in slime and wiring it through --rollout-function-path. Use when user wants to implement custom rollout data generation logic, custom train/eval rollout outputs, or migrate from the default sglang rollout path.

345 2mo ago A 54 tokens

add-tests-and-ci

05

THUDM/DeepDive

Skill Claude CodeCodex

Guide for adding or updating slime tests and CI wiring. Use when tasks require new test cases, CI registration, test matrix updates, or workflow template changes.

345 2mo ago A 36 tokens

THUDM/DeepDive

Skill Claude CodeCodex

Use when reviewing or editing slime code, especially refactors around helper APIs, branch selection, argument validation, or recurring reviewer preferences about avoiding unnecessary wrappers and making control flow self-explanatory.

345 2mo ago A 45 tokens

fix_agent

07

THUDM/SCALE-CUA

Agent

Fix validation errors in generated task configurations and evaluator functions.

55 29d ago A 0 tokens

judge_agent

08

THUDM/SCALE-CUA

Agent

Judge whether a generated verifiable task can be checked correctly by its evaluator. This agent is semantic: rule validators handle schema, imports, and basic signatures; the judge decides whether the evaluator actually measures the instruction.

55 29d ago A 0 tokens

merge_agent

09

THUDM/SCALE-CUA

Agent

Validate and merge generated task files and custom functions into the official directories.

55 29d ago A 0 tokens