vectorspacelab/arex-skill/actor-critic-sil-integration

Claude Code and agent skill by vectorspacelab/arex-skill

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Run a bounded actor-critic Self-Imitation Learning update and log trainable parameter evidence.

Install

npx vskill@latest install vectorspacelab/arex-skill/actor-critic-sil-integration

SKILL.md

The same SKILL.md was published earlier as vectorspacelab/arex-skill/actor-critic-sil-integration.

Actor-Critic SIL Integration

Use this skill when integrating SIL with an actor-critic learner or when building a reduced recovery that must execute the paper's optimizer mechanism. It complements ordinary A2C/PPO updates by applying replay-based positive-advantage imitation updates.

Do not claim this skill reproduces full Atari or MuJoCo training unless the caller uses a real environment, policy network, and training budget.

Inputs

  • Replay records with action and return fields.
  • Initial trainable proxy parameters or a caller-provided actor-critic model.
  • Learning rate, number of updates, and value-loss coefficient.

Outputs

  • Training trace with loss_before, loss_after, params_before, params_after, action probabilities, value estimates, and optimizer-step flags.
  • Mechanism evidence for reduced recovery.

Workflow

  1. Read replay records produced from completed trajectories.
  2. Compute selected-action probability and value estimates from trainable parameters.
  3. Apply SIL positive-advantage policy and value gradients.
  4. Record before/after losses and parameter changes.
  5. Mark reduced/full runtime booleans honestly.

Validation

Run:

bash
python tests/test_scalar_sil_training.py

The test checks that a high-return action receives higher probability and that SIL loss decreases after training.

Limitations

  • The included script is a deterministic scalar proxy for recovery evidence, not a replacement for full A2C/PPO implementations.
  • It intentionally uses only the Python standard library to avoid mutating shared environments.

Bounded Optimizer Sensitivity

Reduced recovery should log the learning rate and update count because very small update budgets may show only tiny probability movement even when parameters change. Acceptance should require parameter-change evidence and a positive loss decrease.

Security

  • Security scanPassed, 90 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/vectorspacelab/arex-skill/actor-critic-sil-integration/badge)](https://verified-skill.com/skills/vectorspacelab/arex-skill/actor-critic-sil-integration)

Something wrong with this skill or its page? Tell us.