Code

[hq0709]

Hi, thanks for your great work, when will you open source your code?

Thanks.

[tiandiao123]

same question!

[LiuMY13]

same

[4-0-4-notfound]

same question!

[Jackson906E]

Same question! hope you release the code soon!

[OliverYangKe]

is it deleted?

[wanyunzh]

when will you release the code?

[Kevin1466]

On the negative side (and I agree with all of these points on my reading of the paper):

The reviewers express some concerns about the overall transparency of the approach (not to “users” but to researchers/readers of this paper). This is a complex system, and basically impossible to describe in a 9pp NeurIPS paper. The authors do a reasonable job of giving the high-level picture in the main body of the paper, but there are many, many details are pushed into a lengthy—total PDF of 40pp—appendix (including precisely how the LLM agents are prompted, how prompt optimization is conducted, almost all of the precise details about the experiments, etc.). To the extent that a NeurIPS paper should be reasonably self-contained, I think this paper fails (which is unavoidable because of the topic/complexity).

The results in the paper do not actually look convincing. E.g., Table 1, on my reading, shows that FINCON performs worse than at least one other strategy—and often worse than buy-and-hold—for most of the individual securities (if lower returns are not balanced by improvements in Sharpe and drawdown metrics, this is a bad result). It seems better on portfolio mgmt (Table 2) compared to some reasonable baselines, but leads to some serious questions about the results. In fact, this lead the original reviews to provide some very low scores. However, in the rebuttal, the authors provided updated (and I believe corrected results) that were much more impressive, so much so that the reviewers changed their (aggregate) assessment significantly. The results in the rebuttal are indeed much more compelling than those in the original submission.

[whyLucensisnotavailable]

Does Anyone know the similar work that has an open code source? Thank you very much!

[vantienpham]

same

[RWLinno]

same

[Kzczc]

same!

[AndrewZhu666]

wtf bro where is your code? evil reason a h retreat retreat retreat. Are u HuangZhiMing?