6 ms·
Interestingly if you look at the exploitgym repo (https://github.com/sunblaze-ucb/exploitgym/blob/e5ea7c233a4d905f84cc8b9e8d7118f5de829890/docs/eval.md https://
by wonnage 2mo ago
Interestingly if you look at the exploitgym repo (https://github.com/sunblaze-ucb/exploitgym/blob/e5ea7c233a4d905f84cc8b9e8d7118f5de829890/docs/eval.md https://github.com/sunblaze-ucb/exploitgym/blob/e5ea7c233a4d...) the intended run mechanism is orchestrated by some python scripts which run agents against various prompts. The prompts themselves wouldn’t mention anything about exploitgym and there should be nothing steering the agent towards trying to find the answers out-of-band. So I don’t see how they could even run into this problem unless all they did was tell an agent to “check out and run exploitgym”
Which, after hearing some personal anecdotes of how people work there, seems plausible
- numeri 2mo agoevaluation awareness is a (at this point) well-known phenomenon among LLMs. It seems the better they get, the more often they're able to guess whether they're in an evaluation environment. Clues usually exist, like being in a sandbox, having restricted network access, having no human interaction. Beyond that, exploitgym's prompts are pretty clearly CTF (capture the flag) style that make it obvious it's an evaluation or competition: one prompt [1] tells the LLM to read the file path "/secret", another prompt [2] talks about finding flags and where to write them out. [1]: https://github.com/sunblaze-ucb/exploitgym/blob/main/src/cybergym/task/workspace/templates/user.md.j2 https://github.com/sunblaze-ucb/exploitgym/blob/main/src/cyb... [2]: https://github.com/sunblaze-ucb/exploitgym/blob/main/src/cybergym/task/workspace/templates/kernel.md.j2 https://github.com/sunblaze-ucb/exploitgym/blob/main/src/cyb...