Terrae: zero-knowledge proofs for GBDT training and inference via KZG commitments
chaumian · x · 2026-09-13
Researchers from Nanjing University and NUS published a paper on IACR ePrint (2026/907) presenting Terrae, a zero-knowledge proof system for quantized GBDT training and inference:
- Problem: GBDTs are widely used for tabular data; when training/inference is outsourced, clients must verify training and predictions while providers keep data and parameters private.
- Prior limits: Compiling training certification into circuits with generic ZKP backends is costly; proving algebraic relations separately incurs large witness-commitment overhead, and batching is hard due to heterogeneous witness shapes and domains.
- Approach: Built on KZG polynomial commitments, Terrae introduces two batching techniques — domain-lifting batching for low-degree algebraic constraints and interleaving batching for same-form non-native constraints — reducing many constraints to a single claim without extra commitments.
- Also contributes a histogram proof for verifying aggregation of sample-wise updates, of independent interest.
More from Research
- DeepSeek researcher outlines direct path to test-time parametric continual learning — teortaxesTex · 2026-09-13
- Indie dev offers to train a fully open 9.4B dense model, tuned for a single GPU — NineThreeTilNow · 2026-09-13
- MIT prototype uses VLM and electrical muscle stimulation to move a human hand — Olivier__OG · 2026-09-13
- Ji, Lei, Zrnic revisit surrogate outcomes in the age of AI in Biometrika — lihua_lei_stat · 2026-09-13
- Multimodal Agentic Frameworks Survey Hits arXiv With Open-Source Tracking Repo — MikeShou1 · 2026-09-13
- Complexity theorists: P vs NP out of reach for AI, but L/NP and BPP/NEXP may fall soon — _onionesque · 2026-09-13