AI Verification Tech MapSearch

Mechanism · Deterministic and bit-exact inference

Sources

On this page
  1. BN. Cankaya (2026). Bit-Exact AI Inference Verification Without Performance Tradeoffs. ICML 2026 Workshop on Technical AI Governance Research. Source recordSupports: covert-adversary threat model; deterministic but non-invariant engines; required metadata; software emulator and its results; limitations; comparison with statistical schemes; best-paper award and public code (arXiv comments) · abstract; §1; results; limitations section; arXiv comments
  2. BR. Rinberg et al. (2025). Verifying LLM Inference to Detect Model Weight Exfiltration. arXiv. Source recordSupports: logging of inferences and random sampling for verification as components separate from recomputation · §4.2, assumptions 2 and 3 (v3)
  3. BN. Cankaya (2026). A System Overview for Near-Term, Low-Trust AI Compute Verification. Machine Intelligence Research Institute. Source recordSupports: replay metadata in a low-trust verification system · §5.2.2
  4. BA. Karvonen et al. (2025). DiFR: Inference Verification Despite Nondeterminism. ICML 2026 Workshop on Technical AI Governance Research. Source recordSupports: sources of benign nondeterminism; fixed-hardware determinism vs heterogeneous deployments; >98% token agreement · §2; §3; §7
  5. CAmodo Design (2026). Example Schemes for Verifying High-Stakes AI Agreements. Amodo Design. Source recordSupports: floating-point non-associativity; fuzzy comparison in recomputation schemes · determinism discussion
  6. CR. Dean (2026). Verification Plan. AI 2040. Source recordSupports: reproducibility required for packet correctness checks · Concrete inference-only retrofitting proposal
  7. CAmodo Design (2026). AI 2040 Plan A — Verification SITREP. Amodo Design. Source recordSupports: status of reproducible inference stack · status items
  8. CH. He & Thinking Machines Lab (2025). Defeating Nondeterminism in LLM Inference. Thinking Machines Lab: Connectionism. Source recordSupports: batch invariance as main cause; kernels made invariant; Qwen3-235B experiment; timings · whole post
  9. AE. Badash et al. (2026). Hawkeye: Reproducing GPU-Level Non-Determinism. Proceedings of Machine Learning and Systems 8 (MLSys 2026). Source recordSupports: exact CPU reproduction of tensor-core matrix multiplication; scope limits · abstract; §8; §9
  10. BDeepSeek-AI (2026). DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence. arXiv. Source recordSupports: provider-reported end-to-end batch-invariant and deterministic kernels · §3.3
  11. CThe SGLang Team (2025). Towards Deterministic Inference in SGLang and Reproducible RL Training. LMSYS Org blog. Source recordSupports: SGLang deterministic mode and its overhead · whole post
  12. BvLLM project (2026). Batch Invariance (vLLM documentation). vLLM documentation (GitHub, docs/features/batch_invariance.md). Source recordSupports: vLLM batch-invariance flag, supported hardware (NVIDIA compute capability 8.0+, Intel XPUs), beta status · whole page
  13. BR. Gond et al. (2026). LLM-42: Enabling Determinism in LLM Inference with Verified Speculation. arXiv. Source recordSupports: scheduling-based determinism alternative that mostly reuses existing kernels · abstract
  14. BThinking Machines Lab (2025). thinking-machines-lab/batch_invariant_ops (GitHub repository). GitHub. Source recordSupports: Thinking Machines' batch-invariant kernel library (MIT) · README
  15. CvLLM project contributors (2025). [Feature]: Batch Invariant Feature and Performance Optimization (vLLM issue #27433). GitHub (vllm-project/vllm issues). Source recordSupports: vLLM developers' statement that batch-invariance support is based on the Thinking Machines post; open work on AMD hardware, NVFP4 and speculative decoding
  16. BA. Arun et al. (2025). Verde: Verification via Refereed Delegation for Machine Learning Programs. arXiv. Source recordSupports: Verde refereed delegation and RepOps reproducible operators · abstract; §3.2
  17. CO. Ersoy (2025). Verde Verification System In Production. Gensyn research blog. Source recordSupports: Gensyn's report of Verde and RepOps in production (provider-reported)
  18. BGensyn (2026). gensyn-ai/ree: Gensyn Reproducible Execution Environment (GitHub repository). GitHub. Source recordSupports: REE public as binaries with an MIT-licensed SDK; v0.8.0 released on 5 October 2026 (provider-reported) · README; patch notes
  19. BD. Ribeiro Alves et al. (2026). EigenAI: Deterministic Inference, Verifiable Results. arXiv. Source recordSupports: EigenAI deterministic engine and optimistic re-execution protocol; same-SKU and A100-vs-H100 determinism results (provider-reported) · abstract; Table 5
  20. CEigenCloud (2025). EigenCloud Brings Verifiable AI to Mass Market with EigenAI and EigenCompute Launches. Eigen Labs blog. Source recordSupports: EigenAI mainnet alpha launch; stake not yet exposed to slashing (provider-reported)
  21. CD. Jedamski (2026). Building Delphi: Pricing, Settlement, and Agentic Trading. Gensyn blog. Source recordSupports: Delphi live on Gensyn's mainnet; REE settlement receipts that anyone can re-run (provider-reported) · settlement section
  22. BGensyn (2026). Reproducible Execution Environment (REE) (Gensyn documentation). Gensyn documentation. Source recordSupports: REE releases carry no alpha or beta label (provider-reported) · whole page
  23. BGensyn (2026). What is Delphi? (Delphi documentation). Delphi documentation. Source recordSupports: Delphi mainnet service and REE judge receipts that anyone can re-run (provider-reported) · Settled by AI, Verifiable by Anyone; Where Delphi Runs
  24. AJ. Yao et al. (2026). TAO: Tolerance-Aware Optimistic Verification for Floating-Point Neural Networks. Proceedings of the 21st European Conference on Computer Systems (EuroSys 2026), pp. 1515-1532. Source recordSupports: TAO: operator-level tolerance bounds instead of bitwise equality on heterogeneous hardware; Merkle-anchored dispute game; Ethereum testnet deployment · abstract; §1; evaluation

Search

Full search page