AuGR: Augmenting Generative-Ranking Joint Training for Chatbot Intent Recommendation

This is the official repository for the AuGR paper. This repository contains code to replicate the results obtained within the paper for public benchmarks, for both sequential recommendation and CTR prediction tasks.

Results on Public Benchmarks

Sequential Recommendation (Amazon 2014)

Only AuGR variants were ran by us, results from other models were obtained from https://phonism.github.io/genrec.

Beauty:

Model	Backbone	R@5	R@10	N@5	N@10
SASRec(CE)	-	0.0538	0.0851	0.0320	0.0421
HSTU (CE)	-	0.0568	0.0859	0.0347	0.0441
TIGER	-	0.0419	0.0644	0.0282	0.0354
LCRec	-	0.0481	0.0704	0.0331	0.0403
OneRec-SFT	-	0.0578	0.0816	0.0398	0.0475
AuGR	TIGER	0.0402	0.0639	0.0260	0.0336
AuGR	HSTU	0.0559	0.0889	0.0355	0.0462

Toys:

Model	Backbone	R@5	R@10	N@5	N@10
SASRec(CE)	-	0.0613	0.0922	0.0348	0.0448
HSTU (CE)	-	0.0611	0.0914	0.0363	0.0461
TIGER	-	0.0340	0.0521	0.0214	0.0272
LCRec	-	0.0433	0.0614	0.0310	0.0368
OneRec-SFT	-	0.0545	0.0790	0.0383	0.0462
AuGR	TIGER	0.0354	0.0557	0.0225	0.0291
AuGR	HSTU	0.0631	0.0934	0.0381	0.0479

Sports:

Model	Backbone	R@5	R@10	N@5	N@10
SASRec(CE)	-	0.0321	0.0495	0.0191	0.0248
HSTU (CE)	-	0.0283	0.0439	0.0182	0.0232
TIGER	-	0.0236	0.0377	0.0150	0.0195
LCRec	-	0.0238	0.0360	0.0159	0.0198
OneRec-SFT	-	0.0299	0.0436	0.0200	0.0244
AuGR	TIGER	0.0251	0.0399	0.0154	0.0201
AuGR	HSTU	0.0326	0.0509	0.0205	0.0265

CTR Prediction (TaoBao Ads)

Model	AUC	LogLoss
SASRec	0.5964	0.2028
AuGR-SASRec	0.5996	0.2060
HSTU	0.5978	0.2013
AuGR-HSTU	0.5994	0.2044

Reproducing the results

Refer to the respective markdowns for the setup instructions.

• CTR Prediction: CTR Prediction README
• Sequential Recommendation: Sequential Recommendation README

References

• https://phonism.github.io/genrec: GenRec: A Model Zoo for Generative Recommendation

Citation

If this repository is useful for your research, please cite the paper:

@misc{augrexperimentcode,
  title = {AuGR: Augmenting Generative-Ranking Joint Training for Chatbot Intent Recommendation},
  author = {Tingting Yu, Roydon Tay, Cheng Chang, Koh Zhi Rong, Bin Fu, Kwan Hui Lim},
  year = {2026},
  eprint={...},
  archivePrefix={arXiv},
  primaryClass={cs.AI},
  url = {...}
}