SEAG-semantic-exploration-with-adaptive-gating

[ACL 2025] Official implementation of Semantic Exploration with Adaptive Gating for Efficient Problem Solving with Language Models

You can visit our project page 🚀!

🚀 Running the GSM8K Experiment

To run the GSM8K experiment using our implementation:

TIKTOKEN_CACHE_DIR="" CUDA_VISIBLE_DEVICES=0 torchrun --nproc-per-node 1 --master-port 1111 \
expample_gsm8k/inference.py \
--base_lm llama-3 --llama_3_ckpts <path_to_llama3_ckpt> --llama_size "8B-Instruct" \
--n_iters 10 --early_term_threshold 11 \
--log_dir ./logs/gsm8k/

• llama_3_ckpts: (Required) Path to the LLaMA 3 checkpoint

Acknowledgements

Our implementation builds upon the following repositories:

• Reasoning with Language Model is Planning with World Model
  → Code: llm-reasoners

• Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation
  → Code: semantic-uncertainty

📬 Contact

If you have any questions, feel free to contact:

• 📧 Hyejin Park: [email protected]
• 📧 Sungjae Lee: [email protected]

Citation

@article{lee2025semantic,
  title={Semantic exploration with adaptive gating for efficient problem solving with language models},
  author={Lee, Sungjae and Park, Hyejin and Kim, Jaechang and Ok, Jungseul},
  journal={arXiv preprint arXiv:2501.05752},
  year={2025}
}