Hui Wei

CS PhD student at UC Merced, previously at UMass Amherst, NYU and NYUMed

[DAIR.AI @dair_ai]

Finally a good paper testing if file-system based memory for LLM agents is worth it.

First, what does this look like?

Deployed agents keep long-term memory as a folder of markdown files they read and reorganize with ordinary file tools.

Two assumptions had never been checked. That an agent can keep a growing store organized as memories accumulate, conflict, and go stale. And whether the organization pays for itself.

Organized stores roughly halve retrieval cost when the material is large. No agent in the study converted organization into better answers, and in the growth study the store degraded for every management agent except the strongest one.

Changing the tool set alone reshapes the memory store as strongly as swapping the model.

Paper: https://t.co/WC6EtdQBfB

Track more trending AI papers in our academy: https://t.co/LRnpZN7L4c

[Hui Wei @HuiWei15]

@feiliu_nlp Have a great time in Vienna 🥳

[Hui Wei @HuiWei15]

@feiliu_nlp Thanks for presenting our PlanGenLLM paper 👏👏

[Hui Wei @HuiWei15]

@chhaviyadav_ @kamalikac Congrats, Chhavi!🥳

[Hui Wei @HuiWei15]

Email: [email protected]
Person Website: https://t.co/r9DdNDgppN
School Website: https://t.co/UFw0efA315

[Hui Wei @HuiWei15]

[PhD Opportunities in Low-Power, Low-Cost, and Ubiquitous Systems Research!!] My friend, Shiwei Fang (https://t.co/r9DdNDgppN), Assistant Professor at Augusta University in Augusta, Georgia, is hiring PhD students in his lab!

[Hui Wei @HuiWei15]

If you are interested in joining his research group, please feel free to contact him via email with your CV and a brief description of your research interests or specific projects you would like to explore.

[Hui Wei @HuiWei15]

@HealthAI_Lab @CHILconference @nutritionorg @ADA_DiabetesPro Thanks for such an awesome talk on CHIL!

[Hui Wei @HuiWei15]

Check out our ICLR paper about contrastive learning on time series data! The work is led by amazing @maxxu05 who will present tomorrow (05/07/2024) at Poster Session 2 @ 4:30 PM as Poster #156!

[Max Xu @maxxu05]

#ICLR2024 How can we choose meaningful positive pairs for time-series contrastive learning? What about motif similarity?
REBAR uses a learned measure that captures motif similarity and achieves SOTA performance.
Arxiv: https://t.co/hvYfgZD1wO
Github: https://t.co/e5oNccvzlz

[Hui Wei @HuiWei15]

@IDeznabi @manningcics Congratulations, Iman!

[Hui Wei @HuiWei15]

What is supposed to do when you find official implementations by the paper authors are different from what they claim in the paper? Trust the paper and change the codes, or just trust the codes and ignore the paper?

[Hui Wei @HuiWei15]

@narges_razavian Thanks for the awesome guidance and being such an excellent advisor, Narges!

[Hui Wei @HuiWei15]

@narges_razavian Congratulations Narges and enjoy your “vacation”!

[Hui Wei @HuiWei15]

Has somebody used advanced learning rate scheduler (e.g. cyclic learning rate, one cycle learning rate) with adaptive learning rate optimizer (e.g. RMSprop, Adam)? If so, how is the performance? Does the scheduler break the internal tracked learning rate in those optimizer? 🧐🧐

[Hui Wei @HuiWei15]

@feiwang03 Congratulations 🎉