Can we combine integer linear programming with exemplar selection to improve In-Context Learning?
Yes! All you need is to optimize your Knapsack 🎒
The paper by
@TongletJ
et al. on
#SEER
was just accepted to
#EMNLP2023
– learn more in this 🧵 (1/8)
📰
1
1
14
Replies
It was observed that the performance of ICL depends heavily on the selection of the exemplars.
@TongletJ
et al. show how this combinatorial optimization problem can be formulated as a Knapsack Integer Linear program and optimized efficiently with deterministic solvers. (2/🧵)
1
0
2
@TongletJ
The Knapsack consists of an objective function, a capacity constraint and optional additional constraints. (3/🧵)
1
0
2
@TongletJ
In their
#EMNLP2023
paper the authors use a capacity constraint to control the size in tokens of the prompt and diversity constraints to favor the selection of exemplars – sharing the same reasoning properties as the test problem. (4/🧵)
1
0
3
@TongletJ
They propose
#SEER
, a method to automatically generate a Knapsack program for
#HybridQA
problems. It achieves superior performance to exemplar selection baselines on the FinQA and TAT-QA datasets. (5/🧵)
#EMNLP2023
1
0
4
@TongletJ
Tokens are the main unit price for commercial LLMs. Thanks to capacity constraints,
#SEER
directly optimizes the prompt size to meet restricted token budgets. (6/🧵)
#EMNLP2023
1
0
3
@TongletJ
If you're interested in our research: We provide open access to our code and results:
➡️
(7/🧵)
#EMNLP2023
1
0
4
@TongletJ
And consider following the authors
@TongletJ
(
@UKPLab
/
@KU_Leuven
), Manon Reusens (
@ManonReusens
), Philipp Borchert (
@IESEG
/
@KU_Leuven
) and
@BartBaesens
(
@KU_Leuven
/
@unisouthampton
), if you are interested in more information or an exchange of ideas. (8/8)
See you in 🇸🇬!
#EMNLP2023
0
1
7