cs.CLNov 11, 2026

VWEN: A Curated Small Language Model for Vietnamese Paraphrasing

Authors: Sang Quang Nguyen

Organizations: University of Information Technology Vietnam National University

Abstract

We present VWEN, a 0.8B-parameter model for Vietnamese paraphrasing at a requested level of change and for five related tasks: generating several paraphrases, simplification, gist, register transfer and paraphrase judgement. VWEN is Qwen3.5-0.8B fine-tuned on about 7k samples chosen by a four-stage pipeline: rules and deduplication, a bidirectional entailment filter, a divergence score that assigns low and high paraphrase levels, and selection for coverage and instruction-following difficulty. We compare it with paraphrase-only recipes from the same pipeline and with training on the ViSP corpus under matched and larger step budgets. On high-divergence paraphrasing, VWEN reaches a success rate (requested level reached and meaning preserved) of 0.61 against 0.46 for the ViSP model at matched compute, a difference that holds on inputs resembling ViSP. Given four times more training examples, the ViSP model closes the gap on such inputs but not on questions, which ViSP lacks and which it often turns into statements. Across all comparisons the entailment filter is the most effective stage. Training on paraphrases alone erases abilities of the base model that the multi-task data preserves.

Figures & tables

Appendix figures & tables8 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

May 30, 2026

Short-form Text Rewriting with Phi Silica

Short-form text rewriting is a constrained variant of paraphrasing in which limited context and high semantic density leave little room for variation. While large language models perform well on general paraphrasing, small language models (SLMs) often struggle with semantic fidelity and hallucination robustness in short-form settings. In this work, we present an empirical study of adapting an SLM, Phi Silica, for short-form rewrite through dataset curation, prompt distillation, parameter-efficient fine-tuning, and evaluation. We curate a dataset of short presentation-style text from public slide decks and use GPT-5-chat both to generate rewrite supervision and to conduct LLM-as-a-judge evaluation. Our results show that finetuning improves semantic fidelity, reduces hallucinations, and increases preference win rate against GPT-5-chat rewrites. The findings suggest that targeted adaptation for SLMs can substantially narrow the gap to cloud models and provide practical guidance for adapting SLMs to precision-critical rewrite tasks.
Jul 2, 2026

BamiBERT: A New BERT-based Language Model for Vietnamese

In this paper, we introduce BamiBERT, a new BERT-based pre-trained language model for Vietnamese that addresses key limitations of PhoBERT -- the current de facto Vietnamese text encoder. Trained from scratch on a 129GB corpus of general-domain Vietnamese text for 20 epochs, BamiBERT supports an extended context length of up to 2048 tokens and operates directly on raw input, eliminating the need for external word segmentation. Across 8 Vietnamese benchmarks, it achieves the best score on 11 of 15 metrics and the second-best on 3 others, setting a new state of the art among "base"-sized Vietnamese encoders and demonstrating strong cross-domain generalization. We release BamiBERT at: https://huggingface.co/Qualcomm-AI-Research/BamiBERT
Sep 24, 2026

R-DEIM Net: An Efficient Rationale-Augmented Dual-Expert Interaction Model for Paraphrase Detection

Recent advances in paraphrase detection reveal a fundamental trade-off: large language models achieve high accuracy but require high computation, while efficient Siamese-BERT variants offer practical scalability with reduced transparency in rationale generation. We present R-DEIM Net, a 76M-parameter dual-expert architecture exploring whether moderate-scale models can achieve competitive accuracy on paraphrase detection while enabling human-readable rationale generation. The architecture combines two specialized components: an Interaction Expert that captures token-level similarity patterns through multi-scale 2D convolutions and attention head allowing variable input length, and a Reasoning Expert that uses a Flan-T5-small decoder to generate rationales as auxiliary supervision. Rather than re-encoding generated text, we extract and pool decoder hidden states as complementary features for classification. On the Quora Question Pairs dataset, R-DEIM Net achieves 90.07% accuracy and 90.16% F1-score via 10-fold cross-validation. This represents competitive performance with strong transformer-based baselines (e.g., MFAE BERT: 90.54% accuracy) and recent large language model based approaches (LLaMA-70B) while using a substantially smaller parameter budget. The model generates rationales alongside predictions, providing potential for auxiliary human-readable descriptions.