Ding et al.
Allocating human feedback by social welfare improves pluralistic alignment
WAPA selects queries for preference labeling based on marginal welfare gains, improving supervision for disadvantaged and priority personas.
Academic
Ruomeng Ding · Tianwei Gao · Lianrui Geng · Mohit Bansal · Zhun Deng
University of North Carolina at Chapel Hill
Research Digest··3 min read
The authors study how to allocate a limited budget of human preference labels across diverse user profiles (personas) when training language models for pluralistic alignment.
Why this paper
From University of North Carolina at Chapel Hill
In one line
Allocating scarce human preference labels by welfare-weighted marginal gain improves pluralistic supervision and downstream preference prediction while controlling which groups benefit.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§