FairProof and ExpProof: certifying ML fairness and explanations with ZKPs
chhaviyadav_ · x · 2026-11-05
The author presents two papers on verifiable ML: FairProof (ICML '24), the first ZKP framework for certifying individual fairness of confidential ML models, made practical by moving expensive computations offline and proving approximate fairness guarantees; and ExpProof (ICML '25), cryptographically verifiable ML explanations that redesign computations to be ZKP-friendly, using efficiently verifiable approximate-optimality guarantees instead of exact solutions. Key lesson: verifiable ML is not just a cryptography problem — it's an ML algorithm design problem.
Related event: FairProof: Verifying ML Fairness with Zero-Knowledge Proofs(2 posts)→
More from Research
- Agent grid eval: center agent scores 1.00, corners drop to 0.46 on map reconstruction — cephaloform · 2026-11-05
- LLM agent grid develops ASCII dialect: "ARE YOU LLM?" emerges at tick 30 — cephaloform · 2026-11-05
- Nine LLM agents in a grid invent their own 128-bit protocol to talk via neighbor buffers — cephaloform · 2026-11-05
- Datalab finds judging errors on 68 of 200 pages in olmOCR-bench, shares 5 checks — VikParuchuri · 2026-11-05
- Datalab tested 15 document-parsing benchmarks: about half misrank parsers vs real errors — VikParuchuri · 2026-11-05
- Stanford AI Lab unveils Joules-to-GDP, a framework measuring economic value per joule of AI compute — KyeGomezB · 2026-11-05