Chenxi Whitehouse

Hi! I am Chenxi, a Research Scientist at Meta FAIR, focusing on Fundamental AI Research for LLMs. I work with Jason Weston in the FAIR Alignment team on post-training and reasoning, leading projects in Reinforcement Learning, LLM-as-a-Judge, and Reward Modeling. I also collaborate with the Llama post-training team to integrate these foundational advances into core models.

Alongside my industry role, I am a visiting researcher at the University of Cambridge, where I was previously a postdoctoral research associate collaborating with Prof. Andreas Vlachos on factuality in NLP. My academic background includes a PhD in knowledge-grounded NLP from City, University of London and a Master’s degree in Electrical Engineering from the University of Erlangen-Nürnberg and University College London.

My current Research Interests include:

Post-training and Reinforcement Learning and Large-Scale Reasoning Models
LLM-as-a-Judge and Generative Reward Modeling
AI Co-Scientist and Co-Improvement

You can find my contact informtion and CV here.

News

Jan - 2026	Two works that I led at Meta are accepted in ICLR 2026! J1: Incentivizing Thinking in LLM-as-a-Judge via Reinforcement Learning MENLO: From Preferences to Proficiency – Evaluating and Modeling Native-like Quality Across 47 Languages!
May - 2025	Two papers from work done at University of Cambridge are accepted at ACL 2025 main conference! What Is That Talk About? A Video-to-Text Summarization Dataset for Scientific Presentations Segment-Level Diffusion: A Framework for Controllable Long-Form Generation with Diffusion Language Models
Oct - 2024	Thrilled to share that I have joined Meta in London as a Research Scientist!
Jul - 2024	Our paper PRobELM: Plausibility Ranking Evaluation for Language Models is accepted in the first Conference of Language Modelling COLM 2024!
Mar - 2024	Our paper Low-Rank Adaptation for Multilingual Summarisation: An Empirical Study from my internship at Google DeepMind is accepted in the findings of NAACL 2024!

Selected Publications

Preprint

Training AI Co-Scientists Using Rubric Rewards

Shashwat Goel, Rishi Hazra, Dulhan Jayalath, Timon Willi, Parag Jain, William F. Shen, Ilias Leontiadis, Francesco Barbieri, Yoram Bachrach, Jonas Geiping, and Chenxi Whitehouse

Abs arXiv PDF

AI co-scientists are emerging as a tool to assist human researchers in achieving their research goals. A crucial feature of these AI co-scientists is the ability to generate a research plan given a set of aims and constraints. The plan may be used by researchers for brainstorming, or may even be implemented after further refinement. However, language models currently struggle to generate research plans that follow all constraints and implicit requirements. In this work, we study how to leverage the vast corpus of existing research papers to train language models that generate better research plans. We build a scalable, diverse training corpus by automatically extracting research goals and goal-specific grading rubrics from papers across several domains. We then train models for research plan generation via reinforcement learning with self-grading. A frozen copy of the initial policy acts as the grader during training, with the rubrics creating a generator-verifier gap that enables improvements without external human supervision. To validate this approach, we conduct a study with human experts for machine learning research goals, spanning 225 hours. The experts prefer plans generated by our finetuned Qwen3-30B-A3B model over the initial model for 70% of research goals, and approve 84% of the automatically extracted goal-specific grading rubrics. To assess generality, we also extend our approach to research goals from medical papers, and new arXiv preprints, evaluating with a jury of frontier models. Our finetuning yields 12-22% relative improvements and significant cross-domain generalization, proving effective even in problem settings like medical research where execution feedback is infeasible. Together, these findings demonstrate the potential of a scalable, automated training recipe as a step towards improving general AI co-scientists.
ICLR

MENLO: From Preferences to Proficiency – Evaluating and Modeling Native-like Quality Across 47 Languages

Chenxi Whitehouse, Sebastian Ruder, Tony Lin, Oksana Kurylo, Haruka Takagi, Janice Lam, Nicolò Busetto, Denise Diaz, and Francisco Guzmán

Abs arXiv PDF

Ensuring native-like quality of large language model (LLM) responses across many languages is challenging. To address this, we introduce MENLO, a framework that operationalizes the evaluation of native-like response quality based on audience design-inspired mechanisms. Using MENLO, we create a dataset of 6,423 human-annotated prompt-response preference pairs covering four quality dimensions with high inter-annotator agreement in 47 language varieties. Our evaluation reveals that zero-shot LLM judges benefit significantly from pairwise evaluation and our structured annotation rubrics, yet they still underperform human annotators on our dataset. We demonstrate substantial improvements through fine-tuning with reinforcement learning, reward shaping, and multi-task learning approaches. Additionally, we show that RL-trained judges can serve as generative reward models to enhance LLMs’ multilingual proficiency, though discrepancies with human judgment remain. Our findings suggest promising directions for scalable multilingual evaluation and preference alignment. We release our dataset and evaluation framework to support further research in multilingual LLM evaluation.
ICLR

J1: Incentivizing Thinking in LLM-as-a-Judge via Reinforcement Learning

Chenxi Whitehouse, Tianlu Wang, Ping Yu, Xian Li, Jason Weston, Ilia Kulikov, and Swarnadeep Saha

Abs arXiv PDF

The progress of AI is bottlenecked by the quality of evaluation, and powerful LLM-as-a-Judge models have proved to be a core solution. Improved judgment ability is enabled by stronger chain-of-thought reasoning, motivating the need to find the best recipes for training such models to think. In this work we introduce J1, a reinforcement learning approach to training such models. Our method converts both verifiable and non-verifiable prompts to judgment tasks with verifiable rewards that incentivize thinking and mitigate judgment bias. In particular, our approach outperforms all other existing 8B or 70B models when trained at those sizes, including models distilled from DeepSeek-R1. J1 also outperforms o1-mini, and even R1 on some benchmarks, despite training a smaller model. We provide analysis and ablations comparing Pairwise-J1 vs Pointwise-J1 models, offline vs online training recipes, reward strategies, seed prompts, and variations in thought length and content. We find that our models make better judgments by learning to outline evaluation criteria, comparing against self-generated reference answers, and re-evaluating the correctness of model responses.
NAACL

Low-Rank Adaptaion for Multilingual Summarisation: An Empirical Study

Chenxi Whitehouse, Fantine Huot, Jasmijn Bastings, Mostafa Dehghani, Chu-Cheng Lin, and Mirella Lapata

Findings of the Association for Computational Linguistics: NAACL 2024

Abs arXiv PDF poster slides

Although the advancements of pre-trained Large Language Models have significantly accelerated recent progress in NLP, their ever-increasing size poses significant challenges for conventional fine-tuning, especially in memory-intensive tasks. We investigate the potential of Parameter-Efficient Fine-Tuning, focusing on Low-Rank Adaptation (LoRA), in the domain of multilingual summarization, a task that is both challenging (due to typically long inputs), and relatively unexplored. We conduct an extensive study across different data availability scenarios, including high- and low-data settings, and cross-lingual transfer, leveraging models of different sizes. Our findings reveal that LoRA is competitive with full fine-tuning when trained with high quantities of data, and excels in low-data scenarios and cross-lingual transfer. We also study different strategies for few-shot cross-lingual transfer, finding that continued LoRA tuning outperforms full fine-tuning and the dynamic composition of language-specific LoRA modules.
EMNLP

LLM-powered Data Augmentation for Enhanced Cross-lingual Performance

Chenxi Whitehouse, Monojit Choudhury, and Alham Fikri Aji

Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing

Abs arXiv PDF code poster slides

This paper explores the potential of leveraging Large Language Models (LLMs) for data augmentation in multilingual commonsense reasoning datasets where the available training data is extremely limited. To achieve this, we utilise several LLMs, namely Dolly-v2, StableVicuna, ChatGPT, and GPT-4, to augment three datasets: XCOPA, XWinograd, and XStoryCloze. Subsequently, we evaluate the effectiveness of fine-tuning smaller multilingual models, mBERT and XLMR, using the synthesised data. We compare the performance of training with data generated in English and target languages, as well as translated English-generated data, revealing the overall advantages of incorporating data generated by LLMs, e.g. a notable 13.4 accuracy score improvement for the best case. Furthermore, we conduct a human evaluation by asking native speakers to assess the naturalness and logical coherence of the generated examples across different languages. The results of the evaluation indicate that LLMs such as ChatGPT and GPT-4 excel at producing natural and coherent text in most languages, however, they struggle to generate meaningful text in certain languages like Tamil. We also observe that ChatGPT falls short in generating plausible alternatives compared to the original dataset, whereas examples from GPT-4 exhibit competitive logical consistency.
ACL

WebIE: Faithful and Robust Information Extraction on the Web

Chenxi Whitehouse, Clara Vania, Alham Fikri Aji, Christos Christodoulopoulos, and Andrea Pierleoni

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics

Abs arXiv PDF code poster

Extracting structured and grounded fact triples from raw text is a fundamental task in Information Extraction (IE). Existing IE datasets are typically collected from Wikipedia articles, using hyperlinks to link entities to the Wikidata knowledge base. However, models trained only on Wikipedia have limitations when applied to web domains, which often contain noisy text or text that does not have any factual information. We present WebIE, the first large-scale, entity-linked closed IE dataset consisting of 1.6M sentences automatically collected from the English Common Crawl corpus. WebIE also includes negative examples, i.e. sentences without fact triples, to better reflect the data on the web. We annotate 21K triples from WebIE through crowdsourcing and introduce mWebIE, a translation of the annotated set in four other languages: French, Spanish, Portuguese, and Hindi. We evaluate the in-domain, out-of-domain, and zero-shot cross-lingual performance of generative IE models and find models trained on WebIE show better generalisability. We also propose three training strategies that use entity linking as an auxiliary task. Our experiments show that adding Entity-Linking objectives improves the faithfulness of our generative IE models.
EACL

Towards a Unified Model for Generating Answers and Explanations in Visual Question Answering

Chenxi Whitehouse, Tillman Weyde, and Pranava Madhyastha

Findings of the Association for Computational Linguistics: EACL 2023

Abs arXiv PDF slides

Providing explanations for visual question answering (VQA) has gained much attention in research. However, most existing systems use separate models for the answers and the explanations. We argue that training explanation models independently of the QA model makes the explanations less grounded and limits performance. To address this, we propose a novel multitask finetuning approach towards a Unified Model for more grounded and consistent generation of both Answers and Explanations (UMAE). To achieve this, we add artificial prompt tokens to training instances and finetune a multimodal encoder-decoder model on various VQA tasks. In our experiments, UMAE models surpass the prior SOTA answer accuracy on A-OKVQA by 10 15%, show competitive results on OK-VQA, achieve new SOTA explanation scores on A-OKVQA and VCR, and demonstrate promising out-of-domain performance on VQA-X.
EMNLP

EntityCS: Improving Zero-Shot Cross-lingual Transfer with Entity-Centric Code Switching

Chenxi Whitehouse, Fenia Christopoulou, and Ignacio Iacobacci

Findings of the Association for Computational Linguistics: EMNLP 2022

Abs arXiv PDF code poster

Accurate alignment between languages is fundamental for improving cross-lingual pre-trained language models (XLMs). Motivated by the natural phenomenon of code-switching (CS) in multilingual speakers, CS has been used as an effective data augmentation method that offers language alignment at word- or phrase-level, in contrast to sentence-level via parallel instances. Existing approaches either use dictionaries or parallel sentences with word-alignment to generate CS data by randomly switching words in a sentence. However, such methods can be suboptimal as dictionaries disregard semantics, and syntax might become invalid after random word switching. In this work, we propose EntityCS, a method that focuses on Entity-level Code-Switching to capture fine-grained cross-lingual semantics without corrupting syntax. We use Wikidata and the English Wikipedia to construct an entity-centric CS corpus by switching entities to their counterparts in other languages. We further propose entity-oriented masking strategies during intermediate model training on the EntityCS corpus for improving entity prediction. Evaluation of the trained models on four entity-centric downstream tasks shows consistent improvements over the baseline with a notable increase of 10% in Fact Retrieval. We release the corpus and models to assist research on code-switching and enriching XLMs with external knowledge.
AAAI-ICWSM

Evaluation of Fake News Detection with Knowledge-Enhanced Language Models

Chenxi Whitehouse, Tillman Weyde, Pranava Madhyastha, and Nikos Komninos

Proceedings of the Sixteenth International AAAI Conference on Web and Social Media

Abs arXiv PDF code poster slides

Recent advances in fake news detection have exploited the success of large-scale pre-trained language models (PLMs). The predominant state-of-the-art approaches are based on fine-tuning PLMs on labelled fake news datasets. However, large-scale PLMs are generally not trained on structured factual data and hence may not possess priors that are grounded in factually accurate knowledge. The use of existing knowledge bases (KBs) with rich human-curated factual information has thus the potential to make fake news detection more effective and robust. In this paper, we investigate the impact of knowledge integration into PLMs for fake news detection. We study several state-of-the-art approaches for knowledge integration, mostly using Wikidata as KB, on two popular fake news datasets - LIAR, a politics-based dataset, and COVID-19, a dataset of messages posted on social media relating to the COVID-19 pandemic. Our experiments show that knowledge-enhanced models can significantly improve fake news detection on LIAR where the KB is relevant and up-to-date. The mixed results on COVID-19 highlight the reliance on stylistic features and the importance of domain specific and current KBs. The code is available at https://github.com/chenxwh/fake-news-detection.