HN
login

realistic

6 stories and discussions about realistic, aggregated from every source we track.

1.
Can open-source prompt-injection detectors catch realistic AI agent attacks?github.com

🛡️ Regex catches 0%, Meta's Prompt Guard 2 catches 1% of 629 realistic AgentDojo injection attacks when they're buried in tool output. Reproducible benchmark. - rudratoshs/buried-injections

6 points•northbridgedev•7 days ago•1 comment•
2.
Anthropic: Emergent Misalignment from Reward Hackinganthropic.com

We show for the first time that realistic AI training processes can accidentally produce misaligned models.

3 points•highfrequency•2 days ago•0 comments•
3.
Generate realistic Postgres test data from DDL, without a databaseweavori.com

Paste your CREATE TABLE statements and generate referentially intact Postgres data. No database, no connection string, no config. Weavori's DDL paste mode runs from a schema file.

2 points•ammarmalik17•3 days ago•0 comments•
4.
Ask HN: Realistic approach to making roon[0] more compact/portable?
1 points•sargstuff•about 1 hour ago•0 comments•
5.
Realistic RCE Exploit Chain in OpenBao and Vaultcontrol-plane.io

ControlPlane and the OpenBao community recently patched a full exploit chain from unauthenticated access to full remote code execution that also affects IBM's HashiCorp Vault.

1 points•cipherboy•1 day ago•1 comment•
6.
Making API Performance Tests More Realistic: From Endpoint Metrics to Role-Based Journeysdev.to

Daniel Balcarek's article API Performance Testing: How to Design Realistic Tests makes a...

0 points•cherware•7 days ago•3 comments

Related topics

testing78chain27metrics5prompt56misalignment6performance90emergent5reward hacking3data290