Papers by Jiajun Shen
Facebook AI’s WAT19 Myanmar-English Translation Task Submission (D19-52)
Copied to clipboard
Peng-Jen Chen, Jiajun Shen, Matthew Le, Vishrav Chaudhary, Ahmed El-Kishky, Guillaume Wenzek, Myle Ott, Marc’Aurelio Ranzato
| Challenge: | Using back-translation, we can improve generalization by using noisy channel re-ranking and ensembling. |
| Approach: | They propose to use BPE-based transformer models to leverage monolingual data to improve generalization and use noisy channel re-ranking and ensembling to improve results. |
| Outcome: | The proposed system improves on the baseline system trained exclusively on the provided small parallel dataset, and the human evaluation and BLEU score are higher. |
The Mirage of Model Editing: Revisiting Evaluation in the Wild (2025.acl-long)
Copied to clipboard
| Challenge: | despite near-perfect results, effectiveness of model editing in real-world applications remains unclear. |
| Approach: | They propose QAEdit and WILD to better reflect real-world use of model editing . they propose a benchmark aligned with widely used question answering datasets and a task-agnostic evaluation framework . |
| Outcome: | The proposed QAEdit benchmark and WILD evaluation framework show that current models perform worse than previously reported. |
Transparentize the Internal and External Knowledge Utilization in LLMs with Trustworthy Citation (2025.findings-acl)
Copied to clipboard
| Challenge: | citation generation and retrieval-augmented generation are still lacking in large language models due to hallucinations. |
| Approach: | They propose a retrieval-augmented citation generation task that requires models to generate citations considering both external and internal knowledge while providing trustworthy references. |
| Outcome: | The proposed method achieves better performance across scenarios compared to baselines . retrieval quality, question types, and model knowledge influence trustworthiness . |
CiteLab: Developing and Diagnosing LLM Citation Generation Workflows via the Human-LLM Interaction (2025.acl-demo)
Copied to clipboard
| Challenge: | Existing frameworks for enabling Large Language Models to generate citations are lacking . however, they can still produce hallucinated responses that are non-factual or irrelevant to the input. |
| Approach: | They propose an open-source and modular framework for enabling LLMs to generate citations in Question-Answering tasks. |
| Outcome: | The proposed framework is extensible and paired with a visual interface, Citefix, facilitating case study and modification of existing citation generation methods. |
VALUE ALIGNMENT TAX: Measuring Value Trade-offs in LLM Alignment (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing work on value alignment characterizes value relations statically, ignoring how interventions reshape the value system. |
| Approach: | They propose a framework that quantifies value trade-offs by measuring how alignment-induced changes propagate across interconnected values relative to achieved on-target gain. |
| Outcome: | The proposed framework measures how value trade-offs propagate across values . it can be used to evaluate intended improvements and unintended side effects . |
The Source-Target Domain Mismatch Problem in Machine Translation (2021.eacl-main)
Copied to clipboard
Jiajun Shen, Peng-Jen Chen, Matthew Le, Junxian He, Jiatao Gu, Myle Ott, Michael Auli, Marc’Aurelio Ranzato
| Challenge: | Despite the interconnected world we live in, people in different places talk about different things in different parts of the world. |
| Approach: | They propose a metric to quantify the effect of local context in machine translation and propose measurable results. |
| Outcome: | The proposed metric can be used to quantify the effect of local context on the use of language in machine translation systems on low resource languages. |
Beyond A Single AI Cluster: A Survey of Decentralized LLM Training (2025.emnlp-main)
Copied to clipboard
| Challenge: | Decentralized LLM training leverages dispersed resources at varying scales. |
| Approach: | They propose a resource-driven paradigm that leverages dispersed resources across clusters, datacenters and even regions. |
| Outcome: | The proposed model scales are 175 billion to 660 billion parameters, and the exponential growth in computational requirements poses significant challenges. |
When is Char Better Than Subword: A Systematic Study of Segmentation Algorithms for Neural Machine Translation (2021.acl-short)
Copied to clipboard
| Challenge: | Subword segmentation algorithms can produce sub-optimal segmentation when the target language is rich in morphological changes or there is not enough data for learning compact composition rules. |
| Approach: | They compare character-based and subword-based neural machine translation systems . they find character-driven models are better at handling morphological phenomena . |
| Outcome: | The character-based models are better at handling morphological phenomena, generating rare and unknown words, and more suitable for transferring to unseen domains. |
Mitigating Object Hallucinations in MLLMs via Multi-Frequency Perturbations (2025.findings-emnlp)
Copied to clipboard
Shuo Li, Jiajun Sun, Guodong Zheng, Xiaoran Fan, Yujiong Shen, Yi Lu, Zhiheng Xi, Yuming Yang, Wenming Tan, Tao Ji, Tao Gui, Qi Zhang, Xuanjing Huang
| Challenge: | Multimodal large language models have demonstrated remarkable performance in visual-language tasks, but their authenticity is often compromised by object hallucinations. |
| Approach: | They propose a multi-frequency perturbation method that leverages both low-frequency and high-frequency features of images to perturb visual feature representations and explicitly suppress redundant frequency-domain features during inference. |
| Outcome: | The proposed method significantly mitigates object hallucinations across various model architectures. |
Embodied-Reasoner: Synergizing Visual Search, Reasoning, and Action for Embodied Interactive Tasks (2026.acl-long)
Copied to clipboard
Wenqi Zhang, Mengna Wang, Gangao Liu, Huixin Xu, Yiwei Jiang, Yongliang Shen, Guiyang Hou, Zhe Zheng, Hang Zhang, Xin Li, Jiajun Liu, Weiming Lu, Peng Li, Yueting Zhuang
| Challenge: | Recent advances in reasoning models have demonstrated remarkable capabilities on mathematical and coding tasks, but their effectiveness in embodied domains remains largely unexplored. |
| Approach: | They propose a reasoning model for interactive embodied tasks that synthesizes 9.3k coherent Observation-Thought-Action trajectories containing 64k ego-centric images and 90k diverse reasoning processes. |
| Outcome: | The proposed model outperforms existing visual reasoning models by +9%, 24%, and +13% on long-horizon tasks. |
The Fall of ROME: Understanding the Collapse of LLMs in Model Editing (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Recent studies have found that model editing methods can cause large language models to collapse with just a single edit. |
| Approach: | They propose a method that uses prefixed keys and adds prefixes during testing to prevent model collapse. |
| Outcome: | The proposed method prevents model collapse while maintaining effectiveness, the authors show . Rank-One Model Editing (ROME) has been found to cause model collapse with just a single edit . |
Beyond Scaling: Measuring and Predicting the Upper Bound of Knowledge Retention in Language Model Pre-Training (2026.acl-long)
Copied to clipboard
Changhao Jiang, Ming Zhang, Yifei Cao, Junjie Ye, Xiaoran Fan, Shihan Dou, Zhiheng Xi, Jiajun Sun, Yi Dong, Yujiong Shen, Jingqi Tong, Baoyu Fan, Tao Gui, Qi Zhang, Xuanjing Huang
| Challenge: | Existing methods to predict performance of large language models are lacking . authors propose a size-dependent mutual information predictor for closed-book question answering accuracy . |
| Approach: | They propose a size-dependent mutual information predictor that integrates knowledge frequency, knowledge specificity, and model size to forecast closed-book question answering accuracy. |
| Outcome: | The proposed method outperforms baseline models and achieves R2 > 0.7 in predicting QA accuracy without additional training. |