LIT @ UMich
- 40 followers
- Ann Arbor, Michigan, USA
- https://lit.eecs.umich.edu
Pinned Loading
Repositories
- kidbench Public
- AtlasNLP Public
AtlasNLP is a resource that maps dataset representation by country and shows that coverage, production, and task diversity are deeply uneven across the global NLP ecosystem.
- hacking-fairness-benchmarks Public
One Example Is Enough to Pass Fairness Benchmarks: Rethinking Fairness Evaluation for Aligned LLMs (EMNLP 2026). Code, data and one-shot GRPO LoRA adapters.
- language-energy-divide Public
The Language–Energy Divide: measuring energy costs of multilingual LLM inference across 122 languages. Analysis code + per-language energy/accuracy results.
- multilingual_reviews_deception Public
Multilingual Deception Detection of GPT-generated Hotel Reviews
- one-shot-grpo-bias Public
🎯 Accepted to COLM 2026 — One flipped example is all it takes: how one-shot GRPO breaks LLM alignment and induces systematic, generalizing bias. Code, data & models for 'It Takes One to Bias Them All.'
- fair-reasoning-steering Public
Wait, am I Being Fair? Characterizing deductive stereotyping in LLM reasoning and mitigating it with Fair-GCG (reasoning-time fairness steering).
- misfired-alignment Public
Code and the VETO benchmark for 'The Wrong Kind of Right: Quantifying and Localizing Misfired Alignment in LLMs'.
- pact_culture_personalization Public
Top languages
Loading…
Most used topics
Loading…