pluralistic-alignment

The open-source repository for PAL: Sample-Efficient Personalized Reward Modeling for Pluralistic Alignment, which provides a general personalized preference learning modeling framework.

A curated collection of papers, benchmarks, datasets, and tools on human values in LLMs and pluralistic alignment.

ICLR 2026 Paper

Reproducibility minimal release — Moral Orientation and Calibration (ICML 2026 Pluralistic Alignment Workshop)

Code for the MTPA framework [EMNLP Findings 2025]

Companion code and data for 'Whose Alignment?' (ICML 2026 Pluralistic Alignment Workshop, arXiv 2605.25256): LLM process alignment on German Credit decisions via the Brunswik lens model

Code of our paper - "VISPA: Pluralistic Alignment via Automatic Value Selection and Activation" (Accepted to EMNLP 2026 (Main))

PEBS: Per-rater Empirical-Bayes Shrinkage for RLHF Reward-Model Calibration (Pluralistic Alignment Workshop @ ICML 2026)

Treating the AI-labeling "ground truth" as a distribution, not a fact — an argument, runnable diagnostics + soft-label/governance export, and four foundational results: social choice · spin glass · topology · information geometry.