EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Hacker News·Aug 31·5 min read
Show HN: Prove your code produced your claims without making reviewers rerun it
TL;DR
A new tool, kveritas, enables tamper-evident verification of computational experiments by binding results to the exact code and hardware used, all in a cryptographically signed PDF. This allows reviewers to trust claims without needing to rerun experiments themselves.
✦ Why It Matters
Start using kveritas today to enhance the credibility of your computational research outputs.
Key Takeaways
- Use Kveritas to bind results to code and hardware for verification.
- Generate a signed PDF report to prove your computational claims.
- Leverage selective disclosure to protect sensitive data.
- Capture detailed telemetry data during experiments for robust validation.
How It Works
Kveritas operates by recording telemetry data during the execution of computational experiments, including CPU and memory usage. It generates a signed timeline of snapshots that captures the state of the system at various phases of execution.
This data is then used to verify the coherence of the results, ensuring that they are genuine and not fabricated. The tool employs cryptographic techniques to bind the results to the specific execution context, making it impossible to alter the data without detection.