Skip to content

Conversation

@hflyzju
Copy link

@hflyzju hflyzju commented Jul 11, 2025

No description provided.

hflyzju and others added 22 commits July 11, 2025 08:33
Add support for google/gemini-2.5-flash-preview-05-20 model
Update API configuration to use OpenRouter instead of direct Gemini API
Update README with new model usage example
Add explanatory comments to clarify the purpose of coder.run and run_experiment calls
…ration

Add new prompt template and logic to incorporate literature review results into research idea generation process. Also includes saving new ideas separately to track literature-influenced ideas.
- Add nova index API integration for paper search
- Enhance generate_ideas with nova index and semantic index options
- Update LLM client creation for deepseek and azure models
- Add use-literature flag and lit-review-size parameter
- Improve error handling and logging in literature review
- Add test files for aider integration
Add new --skip-run-experiment flag to skip experiment execution
Uncomment and enable saving generated ideas to JSON files
Add clarification in lit review prompt to avoid copying ideas
feat: add skip-run-experiment flag and enable idea saving
updated the code and paper write
Add debug flag to limit novel ideas processing and print debug info
Update ideas file name from ideas.json to new_ideas.json
Add new example command with debug flags to README
fix(plot): handle empty arrays gracefully in loss calculation
refactor(launch): add skip-write-paper flag and engine parameter
style(experiment): disable torch compile by default
… version)

feat(greeting): add name argument support and improve script structure
refactor(test): update test files and add new test utilities
docs: update README with new experiment commands
fix(writeup): improve citation handling and paper structure
… generation

- Add launch_scientist_exp2.py for running AI scientist experiments
- Create run_exp2_0719.sh script to execute experiment2
- Update README.md with experiment2 instructions
- Add generate_ideas_after_paper_review.py for idea generation with literature review
…usion models

Add PDF reports, training/validation loss plots, and performance metrics for various experiments including hidden state hierarchy, ngram to word emergence, attention evolution analysis, and diffusion guidance techniques. Includes JSON files with final training metrics for each run.
Add launch script for experiment1 with deepseek-chat model
Include experiment1 commands in README
Implement generate_ideas_with_proposal_review module
… format

- Update prompt template to focus on generating detailed proposals with response letters
- Remove unused prev_ideas_string parameter
- Change output filename to include version indicator
- Clean up commented debug code
refactor(perform_review): clean up whitespace and improve prompt structure
docs(readme): update with experiment2_v2 commands
chore(gitignore): add new experiment directories
Add script to analyze rating changes between paper versions v1 and v2
Generate GSB distribution and metrics comparison plots
Save rating changes and paper changes list to csv files
- Add launch_scientist_exp2_v4.py with improvements over v3
- Add perform_review_v4.py for enhanced paper review functionality
- Update gitignore to include new experiment result directories
- Add run_exp2_v4_0728.sh script for batch execution
- Implement collect_final_result.py for result aggregation
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants