Most | Son The Nguyen

Hands-on tutorials on aligning language models at inference time without modifying model weights.

By Son The Nguyen, Tommy Cheng, and Theja Tulabandhula.

Project resources

GitHub Repository Archived v1.0.0 Project DOI NAIRR 2026 Recorded Tutorial

If you use the Training-Free Alignment tutorials or notebooks in your work, please cite:

@software{Nguyen_Training-Free_Alignment_Hands-on_2026,
  author = {Nguyen, Son The and Cheng, Tommy and Tulabandhula, Theja},
  doi = {10.5281/zenodo.21925779},
  month = mar,
  title = {{Training-Free Alignment: Hands-on Tutorials on Aligning Language Models at Inference Time}},
  url = {https://github.com/sonthenguyen/alignment},
  version = {1.0.0},
  year = {2026}
}

Browse by topic

Choose a topic to filter the posts shown.

03 August 11, 2026 Response Engineering for Training-Free Alignment of LLMs A practical tutorial on Best-of-N, self-consistency, Reflexion, and Mixture-of-Agents for improving model responses without retraining.
02 August 11, 2026 Guided Decoding for Training-Free Alignment of LLMs A hands-on tutorial covering standard decoding strategies and guided decoding with logits processors.
01 August 11, 2026 Prompt Engineering for Training-Free Alignment of LLMs A hands-on DSPy tutorial for programmatic prompt design, evaluation, and optimization with MIPROv2.