Learning to Explore and Select for Coverage-Conditioned Retrieval-Augmented Generation
Authors:
Takyoung Kim,
Kyungjae Lee,
Young Rok Jang,
Ji Yong Cho,
Gangwoo Kim,
Minseok Cho,
Moontae Lee
Abstract:
Interactions with billion-scale large language models typically yield long-form responses due to their extensive parametric capacities, along with retrieval-augmented features. While detailed responses provide insightful viewpoint of a specific subject, they frequently generate redundant and less engaging content that does not meet user interests. In this work, we focus on the role of query outlin…
▽ More
Interactions with billion-scale large language models typically yield long-form responses due to their extensive parametric capacities, along with retrieval-augmented features. While detailed responses provide insightful viewpoint of a specific subject, they frequently generate redundant and less engaging content that does not meet user interests. In this work, we focus on the role of query outlining (i.e., selected sequence of queries) in scenarios that users request a specific range of information, namely coverage-conditioned ($C^2$) scenarios. For simulating $C^2$ scenarios, we construct QTree, 10K sets of information-seeking queries decomposed with various perspectives on certain topics. By utilizing QTree, we train QPlanner, a 7B language model generating customized query outlines that follow coverage-conditioned queries. We analyze the effectiveness of generated outlines through automatic and human evaluation, targeting on retrieval-augmented generation (RAG). Moreover, the experimental results demonstrate that QPlanner with alignment training can further provide outlines satisfying diverse user interests. Our resources are available at https://github.com/youngerous/qtree.
△ Less
Submitted 1 July, 2024;
originally announced July 2024.
Pseudogap of Superconducting correlation origin in iron-pnictide superconductor Ba0.6+δK0.4-δFe2As2
Authors:
Yong Seung Kwon,
Jong Beom Hong,
Yu Ran Jang,
Hyun ** Oh,
Yun Young Song,
Byeong Hun Min,
Takuya Iizuka,
Shin-ichi Kimura,
A. V. Balatsky,
Yunkyu Bang
Abstract:
Pseudogap (PG), a phenomenon of opening of gap like features above superconducting (SC) transition temperature (Tc), has been universally observed in the high-Tc cuprates (HTC) (1,2), some heavy fermion superconductors (3), and iron pnictides (4-7). Here, we report the observation of PG behavior in optical conductivity in an underdoped Ba0.6+δK0.4-δFe2As2 (Tconset =36 K) single crystal far above T…
▽ More
Pseudogap (PG), a phenomenon of opening of gap like features above superconducting (SC) transition temperature (Tc), has been universally observed in the high-Tc cuprates (HTC) (1,2), some heavy fermion superconductors (3), and iron pnictides (4-7). Here, we report the observation of PG behavior in optical conductivity in an underdoped Ba0.6+δK0.4-δFe2As2 (Tconset =36 K) single crystal far above Tc, up to 100 K (~ 3Tc). Unique scale separation in Ba0.6+δK0.4-δFe2As2 compound - magnetic and SC correlation energy scales being widely separated - enabled us to establish that the PG structures observed in the range of 50 cm-1 - 150 cm-1 in optical conductivity is caused by the SC correlation and the magnetic correlation origin is ruled out. Theoretical calculations, based on the preformed Cooper pair model (8), provided an excellent description to the temperature evolution of the optical conductivity data from below to above Tc.
△ Less
Submitted 21 July, 2010;
originally announced July 2010.
