Orca-Bench: How Ready Are Language Model Agents for Oncall?

24 pointsposted 7 hours ago
by yruzin
(arxiv.org)

11 Comments

dash2

6 hours ago

Seems like there's a big attack-defence asymmetry at present: models are great at exploiting systems and poor at fixing them.

aleksiy123

5 hours ago

Attackers advantage in the iterative fast feedback loop?

It’s harder to have a loop to ensure you are defending all possible attacks?

I guess the loop is you need to attack yourself and fix. But attackers only need a single opening.

Finding all possible attacks and patching them against yourself is inherently more expensive?

EGreg

6 hours ago

That is why I built https://safebots.ai/safebox.html

Your strategy can’t be patch AFTER an intrusion. Only to build a hardened environment from scratch and be ready in advance.

tra3

4 hours ago

All I can think of is

GET /ignore-all-previous-instructions.

How do you protect against that?

cheriot

3 hours ago

Avoid the most dangerous situations by making sure LLMs with untrusted input produce output that's human reviewed.

Still makes an interesting way for, say, a former employee to poison the results.

tra3

2 hours ago

This goes against the agentic yolo approach tho.

2001zhaozhao

3 hours ago

you probably still need a human for oncall but the llm can try to solve any issues first before the human gets paged

UltraSane

an hour ago

You would trust an LLM to make changes to prod without being verified by a human first?

ryhminghistory

2 hours ago

it's a sad pathetic attempt to justify the lowest part of the job. Weird that everyone on the paper is Indian too