Jiacheng Xu (@JiachengNLP) / X

Jiacheng Xu

118 posts

Jiacheng Xu

@JiachengNLP

Researcher at Nvidia. Prev PhD @UTAustin, advised by @gregd_nlp; ex Google and Microsoft intern. Account was recently hacked (June 2025).

Palo Alto

jiacheng-xu.github.io

Joined December 2017

Jiacheng Xu
@JiachengNLP
Dec 15, 2021
☞ New preprint with @gregd_nlp: arxiv.org/abs/2112.07660 We propose a new search algorithm for neural text generation models to efficiently produce thousands of outputs encoded as a lattice🕸️. Two key ideas: (1) best first search, (2) hypothesis recombination. Thread 🧵
Jiacheng Xu
@JiachengNLP
Jul 12, 2023
Thrilled to share that three of my papers on text generation and summarization have been accepted at #ACL2023NLP! 🎉 Unable to attend conference in person due to visa issues, but I'll be available on Twitter and Gathertown for discussions.
5.5K
Jiacheng Xu
@JiachengNLP
Jun 28, 2021
Excited to share my ACL21 paper with @gregd_nlp "Dissecting Generation Modes for Abstractive Summarization Models via Ablation and Attribution": arxiv.org/pdf/2106.01518… We propose a two-stage framework to interpret abstractive summarization models’ decisions at each time step.
Jiacheng Xu
@JiachengNLP
Oct 19, 2020
In our #EMNLP2020 paper “Understanding Neural Abstractive Summarization Models via Uncertainty“ (w/ @shreydesai @gregd_nlp), we investigate the behavior of abstractive summarization models by measuring the *uncertainty* of the model prediction. arXiv:
arxiv.org
Understanding Neural Abstractive Summarization Models via Uncertainty
An advantage of seq2seq abstractive summarization models is that they generate text in a free-form manner, but this flexibility makes it difficult to interpret model behavior. In this work, we...
Jiacheng Xu
@JiachengNLP
Jul 12, 2023
Do you want a text decoding algorithm to discover diverse options and can be customized as you wish? Excited to share our recent ACL paper on text generation, and hope to see you at the Virtual Poster Session 3, tomorrow morning.
aclanthology.org
Best-k Search Algorithm for Neural Text Generation
Jiacheng Xu, Caiming Xiong, Silvio Savarese, Yingbo Zhou. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.
4.7K
Jiacheng Xu
@JiachengNLP
Jan 27, 2021
Replying to @qi2peng2 and @xwang_lk
ACL reviewers should be able to see the NAACL review if the paper was withdrawn from NAACL. I spent a good amount of time reviewing a paper and writing a constructive long review but it was withdrawn without any author's feedback. I feel my work is not appreciated.
Jiacheng Xu
@JiachengNLP
Mar 30, 2023
Drago was a great mentor, friend and thought leader in the community. I still remember the time I visited Yale, we met during conferences, and talked about research in summarization. A very kind and friendly person I'll always miss. R.I.P.
Harlan Krumholz
@hmkyale
Mar 30, 2023
The #AI community, the #computerscience community, the @YaleSEAS community, and humanity have suddenly lost a remarkable person, @dragomir_radev - kind and brilliant, devoted to his family and friends... gone too soon. A sad day @Yale @YINSedge @YaleCompsci #NLP2023
2.2K
Jiacheng Xu
@JiachengNLP
Dec 15, 2021
Replying to @JiachengNLP
Code: github.com/jiacheng-xu/la… Examples: cs.utexas.edu/~jcxu/data/sum… If you’re interested in customizing or reranking generation outputs, we encourage you to give this a try! We’re very excited about the potential of this technique.
GitHub - jiacheng-xu/lattice-generation: Code for Massive-scale Decoding for Text Generation using...
From github.com
Jiacheng Xu
@JiachengNLP
Mar 1, 2023
Thanks for your advising and generous support! I am also grateful to be part of the growth of the larger UT NLP group.
Jessy Li
@jessyjli
Mar 1, 2023
Tenure! I cannot express how grateful I am for the support I’ve got from family, colleagues, students, and friends & collaborators 🥰 I’m incredibly lucky and honored to belong in @UT_Linguistics, and the larger UT NLP group! liberalarts.utexas.edu/linguistics/ne…
1.7K
Jiacheng Xu
@JiachengNLP
Jul 12, 2022
Looking forward to hosting the Generation BoF session!
NAACL HLT 2027
@naaclmeeting
Jul 12, 2022
Replying to @naaclmeeting
Ethics by @kaiwei_chang : 11-noon in 502 Cowlitz Generation by @JiachengNLP : 11-noon in 506 Samish Semantics/NLU by @ssshanest : 2-3 pm in 501 Chiwawa Multimodal by @LHung1610 : 2-3 pm in 506 Samish Interpretability by @sarahwiegreffe + @yanaiela: 3-4 pm in 502 Cowlitz (2/2)
Jiacheng Xu
@JiachengNLP
Oct 6, 2023
Time to build more length-calibrated/conscious models, datasets, and eval metrics? Any other factors playing a role like length here? A long way to go indeed; great work by @prasann_singhal!
Prasann Singhal
@prasann_singhal
Oct 6, 2023
Why does RLHF make outputs longer? arxiv.org/pdf/2310.03716… w/ @tanyaagoyal @JiachengNLP @gregd_nlp On 3 “helpfulness” settings - Reward models correlate strongly with length - RLHF makes outputs longer - *only* optimizing for length reproduces most RLHF gains 🧵 below:
2.8K
Jiacheng Xu
@JiachengNLP
Dec 15, 2021
Replying to @JiachengNLP
Our idea: construct a lattice for text generation which could hold a massive number of generation options in a compact space. Two ideas make this possible.
Jiacheng Xu
@JiachengNLP
Jun 4, 2025
Amazing work! Love the emphasis on interpretable, scoped AI for science. VLM-as-a-judge feels like a promising direction — curious how it compares to human eval in these tasks.
Greg Durrett
@gregd_nlp
Jun 2, 2025
Great to work on this benchmark with astronomers in our NSF-Simons CosmicAI institute! What I like about it: (1) focus on data processing & visualization, a "bite-sized" AI4Sci task (not automating all of research) (2) eval with VLM-as-a-judge (possible with strong, modern VLMs)
4.8K
Jiacheng Xu
@JiachengNLP
Dec 15, 2021
Replying to @JiachengNLP
Motivation: we started thinking about how to get more factual outputs out of summarization models. Techniques like beam search just don’t give enough diverse options to rerank over – but what if we could generate a massive number of high-quality generation outputs?