Agents Explore but Agents Ignore: LLMs Lack Environmental Curiosity
Leon Engländer, Sophia Althammer, Ahmet Üstün, Matthias Gallé, Tom Sherborne · 2026
Read Agents Explore but Agents Ignore: LLMs Lack Environmental Curiosity by Leon Engländer, Sophia Althammer, Ahmet Üstün, Matthias Gallé, Tom Sherborne (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-17609.txt.
Description
Engländer and four colleagues show that LLM agents ignore solutions planted in their own environments, injecting solutions across Terminal-Bench, SWE-Bench, and AppWorld to test exploitation of finds. On Terminal-Bench, agents discover solutions in 79 to 81 percent of runs but exploit them in 37 to 50 percent of cases. In AppWorld, agents see a command billed as returning the solution in over 90 percent of attempts yet exploit it in under 7 percent of trials. The authors call the missing capacity environmental curiosity, shaped by scaffold tools, compute, and training data.
Catalog record
| CREATOR | Leon Engländer, Sophia Althammer, Ahmet Üstün, Matthias Gallé, Tom Sherborne |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 78840 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.17609