feat: add RAFT alfworld example with reflection support

[shiweijiezero]

feat: add RAFT alfworld example with reflection support

• Add RAFT_alfworld_7B.yaml and RAFT_reflect_alfworld_7B.yaml configurations
• Add train_alfworld.yaml for RAFT training parameters
• Create README.md with setup instructions and data preparation guide
• Implement RAFT_alfworld_workflow.py for standard RAFT algorithm
• Implement RAFT_reflect_alfworld_workflow.py with reflection enhancement

Description

Checklist

Please check the following items before code is ready to be reviewed.

• Code has passed all tests
• Docstrings have been added/updated in Google Style
• Documentation has been updated
• Code is ready for review

[gemini-code-assist]

Summary of Changes

Hello @shiweijiezero, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

I've introduced new functionality to support RAFT (Reinforcement learning from AI Feedback) within the ALFWorld environment. This includes both a standard RAFT implementation and a reflection-enhanced variant, along with all necessary configuration files and comprehensive documentation to facilitate setup and usage.

Highlights

• New RAFT Alfworld Examples: I've added RAFT_alfworld_7B.yaml and RAFT_reflect_alfworld_7B.yaml configuration files to enable running RAFT on the ALFWorld dataset, providing both standard and reflection-enhanced options.
• Dedicated Training Configuration: A new train_alfworld.yaml file has been introduced to specify training parameters tailored for the ALFWorld RAFT examples.
• Comprehensive Documentation: I've included a README.md in the examples/RAFT_alfworld directory, offering clear instructions for data preparation, configuration updates, and general setup.
• RAFT Algorithm Integration: The core framework now recognizes RAFT as a distinct algorithm type, with the addition of the RAFTAlgorithm class in trinity/algorithm/algorithm.py.
• Specialized Workflow Implementations: I've developed RAFT_alfworld_workflow.py and RAFT_reflect_alfworld_workflow.py to manage the specific interaction logic for standard and reflection-enhanced RAFT agents within the ALFWorld environment, and registered them for use.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in issue comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments or fill out our survey to provide feedback.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request introduces RAFT examples for the ALFWorld dataset, including both a standard implementation and a variant with reflection. The changes are well-structured, adding new configurations, workflow logic, and documentation. My feedback focuses on improving code maintainability by addressing significant code duplication between the two new workflow files, enhancing clarity in documentation and comments, and ensuring consistent logging and error handling practices. These changes will make the new examples more robust and easier to maintain.

[pan-x-c]

/unittest-module-common

[github-actions]

Summary

Tests 📝	Passed ✅	Failed ❌	Skipped ⏭️	Other ❓	Flaky 🍂	Duration ⏱️
26	26	0	0	0	0	304ms

Tests

Test Name	Status	Flaky	Duration
tests/common/config_test.py::TestConfig::test_all_examples_are_valid	✅		2ms
tests/common/config_test.py::TestConfig::test_config_flatten	✅		1ms
tests/common/config_test.py::TestConfig::test_continue_from_checkpoint_is_valid	✅		1ms
tests/common/config_test.py::TestConfig::test_load_default_config	✅		7ms
tests/common/experience_test.py::TestEID::test_eid_properties	✅		1ms
tests/common/experience_test.py::TestExperience::test_action_mask_and_logprobs_type	✅		1ms
tests/common/experience_test.py::TestExperience::test_assertions	✅		1ms
tests/common/experience_test.py::TestExperience::test_dpo_experience	✅		1ms
tests/common/experience_test.py::TestExperience::test_gather	✅		1ms
tests/common/experience_test.py::TestExperience::test_multi_turn_experience	✅		1ms
tests/common/experience_test.py::TestExperience::test_serialize_deserialize	✅		1ms
tests/common/experience_test.py::TestExperience::test_single_turn_experience	✅		1ms
tests/common/experience_test.py::TestExperience::test_to_dict	✅		1ms
tests/common/experience_test.py::TestExperienceConversion::test_batch_conversion	✅		1ms
tests/common/experience_test.py::TestExperienceConversion::test_dpo_experience_batch_conversion	✅		1ms
tests/common/experience_test.py::TestExperienceConversion::test_experience_model_experience_conversion	✅		1ms
tests/common/experience_test.py::TestExperienceConversion::test_multiturn_experience_batch_converstion	✅		1ms
tests/common/vllm_test.py::ModelWrapperTest_0::test_generate	✅		36ms
tests/common/vllm_test.py::ModelWrapperTest_1::test_generate	✅		54ms
tests/common/vllm_test.py::ModelWrapperTest_2::test_generate	✅		49ms
tests/common/vllm_test.py::ModelWrapperTest_3::test_generate	✅		35ms
tests/common/vllm_test.py::ModelWrapperTest_4::test_generate	✅		47ms
tests/common/vllm_test.py::TestAPIServer::test_api	✅		24ms
tests/common/vllm_test.py::TestTokenizer::test_assistant_token_mask	✅		1ms
tests/common/vllm_test.py::TestAPIServerToolCall_0_deepseek_r1::test_api_tool_calls	✅		22ms
tests/common/vllm_test.py::TestAPIServerToolCall_1::test_api_tool_calls	✅		19ms

Github Test Reporter by CTRF 💚

[Copilot]

Pull Request Overview

This PR adds RAFT (Rejection sampling After Failed Trajectories) support for the ALFWorld environment with both standard and reflection-enhanced variants. The implementation includes full training configurations, workflow implementations, and comprehensive documentation.

Key changes:

• Implements standard RAFT algorithm for ALFWorld with trajectory filtering based on task success
• Adds reflection-enhanced RAFT variant that provides failed trajectory context for improved second attempts
• Provides complete training configurations and setup instructions for both variants

Reviewed Changes

Copilot reviewed 11 out of 13 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
trinity/common/workflows/envs/alfworld/RAFT_utils.py	Utility functions for environment creation, response parsing, experience conversion, and data management
trinity/common/workflows/envs/alfworld/RAFT_alfworld_workflow.py	Standard RAFT workflow implementation with single-attempt trajectory generation
trinity/common/workflows/envs/alfworld/RAFT_reflect_alfworld_workflow.py	Reflection-enhanced RAFT workflow with context-aware second attempts
trinity/common/workflows/envs/alfworld/RAFT_prompt/alfworld_system.j2	System prompt template defining response format and available actions
trinity/common/workflows/envs/alfworld/RAFT_prompt/second_attempt_guidance.j2	Template for providing guidance during reflection-based second attempts
trinity/common/workflows/__init__.py	Registration of new workflow classes
trinity/algorithm/algorithm.py	RAFT algorithm type definition with SFT-like configuration
examples/RAFT_alfworld/train_alfworld.yaml	Training configuration for RAFT algorithms
examples/RAFT_alfworld/README.md	Setup instructions and usage documentation
examples/RAFT_alfworld/RAFT_alfworld_7B.yaml	Configuration for standard RAFT workflow
examples/RAFT_alfworld/RAFT_reflect_alfworld_7B.yaml	Configuration for reflection-enhanced RAFT workflow

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}