Arnav Goel

Arnav Goel, Pranjal A Chitale, Bhawna Paliwal et al. · microsoft-research

User behavior in the real world is diverse, cross-domain, and spans long time horizons. Existing user modeling benchmarks however remain narrow, focusing mainly on short sessions and next-item prediction within a single domain. Such limitations hinder progress toward robust and generalizable user models. We present HORIZON, a new benchmark that reformulates user modeling along three axes i.e. dataset, task, and evaluation. Built from a large-scale, cross-domain reformulation of Amazon Reviews, HORIZON covers 54M users and 35M items, enabling both pretraining and realistic evaluation of models in heterogeneous environments. Unlike prior benchmarks, it challenges models to generalize across domains, users, and time, moving beyond standard missing-positive prediction in the same domain. We propose new tasks and evaluation setups that better reflect real-world deployment scenarios. These include temporal generalization, sequence-length variation, and modeling unseen users, with metrics designed to assess general user behavior understanding rather than isolated next-item prediction. We benchmark popular sequential recommendation architectures alongside LLM-based baselines that leverage long-term interaction histories. Our results highlight the gap between current methods and the demands of real-world user modeling, while establishing HORIZON as a foundation for research on temporally robust, cross-domain, and general-purpose user models.

Arnav Goel, Medha Hira, Avinash Anand et al.

The previous work on controllable text generation is organized using a new schema we provide in this study. Seven components make up the schema, and each one is crucial to the creation process. To accomplish controlled generation for scientific literature, we describe the various modulation strategies utilised to modulate each of the seven components. We also offer a theoretical study and qualitative examination of these methods. This insight makes possible new architectures based on combinations of these components. Future research will compare these methods empirically to learn more about their strengths and utility.

Mehul Agarwal, Aditya Aggarwal, Arnav Goel et al.

While multilingual large language models (LLMs) perform well on high-level tasks like translation and question answering, their ability to handle grammatical gender and morphological agreement remains underexplored. In morphologically rich languages, gender influences verb conjugation, pronouns, and even first-person constructions with explicit and implicit mentions of gender. We introduce MORPHOGEN, a morphologically grounded large-scale benchmark dataset for evaluating gender-aware generation in three typologically diverse grammatically gendered languages: French, Arabic, and Hindi. The core task, GENFORM, requires models to rewrite a first-person sentence in the opposite gender while preserving its meaning and structure. We construct a high-quality synthetic dataset spanning these three languages and benchmark 15 popular multilingual LLMs (2B-70B) on their ability to perform this transformation. Our results reveal significant gaps and interesting insights into how current models handle morphological gender. MORPHOGEN provides a focused diagnostic lens for gender-aware language modeling and lays the groundwork for future research on inclusive and morphology-sensitive NLP.

Shubh Goyal, Medha Hira, Shubham Mishra et al.

Although the rise of Large Language Models (LLMs) in enterprise settings brings new opportunities and capabilities, it also brings challenges, such as the risk of generating inappropriate, biased, or misleading content that violates regulations and can have legal concerns. To alleviate this, we present "LLMGuard", a tool that monitors user interactions with an LLM application and flags content against specific behaviours or conversation topics. To do this robustly, LLMGuard employs an ensemble of detectors.

Huihan Li, Arnav Goel, Keyu He et al.

In open-ended generative tasks like narrative writing or dialogue, large language models often exhibit cultural biases, showing limited knowledge and generating templated outputs for less prevalent cultures. Recent works show that these biases may stem from uneven cultural representation in pretraining corpora. This work investigates how pretraining leads to biased culture-conditioned generations by analyzing how models associate entities with cultures based on pretraining data patterns. We propose the MEMOed framework (MEMOrization from pretraining document) to determine whether a generation for a culture arises from memorization. Using MEMOed on culture-conditioned generations about food and clothing for 110 cultures, we find that high-frequency cultures in pretraining data yield more generations with memorized symbols, while some low-frequency cultures produce none. Additionally, the model favors generating entities with extraordinarily high frequency regardless of the conditioned culture, reflecting biases toward frequent pretraining terms irrespective of relevance. We hope that the MEMOed framework and our insights will inspire more works on attributing model performance on pretraining data.

Medha Hira, Arnav Goel, Anubha Gupta

This paper presents CrossVoice, a novel cascade-based Speech-to-Speech Translation (S2ST) system employing advanced ASR, MT, and TTS technologies with cross-lingual prosody preservation through transfer learning. We conducted comprehensive experiments comparing CrossVoice with direct-S2ST systems, showing improved BLEU scores on tasks such as Fisher Es-En, VoxPopuli Fr-En and prosody preservation on benchmark datasets CVSS-T and IndicTTS. With an average mean opinion score of 3.75 out of 4, speech synthesized by CrossVoice closely rivals human speech on the benchmark, highlighting the efficacy of cascade-based systems and transfer learning in multilingual S2ST with prosody transfer.

Arnav Goel, Medha Hira, Anubha Gupta

The field of prosody transfer in speech synthesis systems is rapidly advancing. This research is focused on evaluating learning methods for adapting pre-trained monolingual text-to-speech (TTS) models to multilingual conditions, i.e., Supervised Fine-Tuning (SFT) and Transfer Learning (TL). This comparison utilizes three distinct metrics: Mean Opinion Score (MOS), Recognition Accuracy (RA), and Mel Cepstral Distortion (MCD). Results demonstrate that, in comparison to SFT, TL leads to significantly enhanced performance, with an average MOS higher by 1.53 points, a 37.5% increase in RA, and approximately a 7.8-point improvement in MCD. These findings are instrumental in helping build TTS models for low-resource languages.

Janak Kapuriya, Anwar Shaikh, Arnav Goel et al.

In this study, we introduce Vision-Caption aware Supervised FineTuning (VCASFT), a novel learning paradigm designed to enhance the performance of smaller Vision Language Models(VLMs) on scientific visual question answering(VQA) tasks. VCASFT leverages image captions as zero-shot prompts alongside question-answer pairs and instruction-tunes models to yield significant performance improvements. To comprehensively evaluate VCASFT, we benchmark it on ScienceQA, which consists of questions across diverse languages, subjects, and fields, demonstrating its adaptability and effectiveness in a variety of educational contexts. Additionally, to further demonstrate the effectiveness of this technique on lowresource languages, we developed HiSciVQA, a dataset comprising 2,245 high-quality, hand-annotated Hindi multimodal Q&A pairs. This dataset addresses the critical need for low-resource language Q&A datasets and serves as a foundation for testing VCASFT. Additionally, we introduce a novel LLM-based evaluation scheme to evaluate VLMs on HiSciVQA which offers deeper insights into model effectiveness surpassing traditional n-gram matching accuracy metrics. We are committed to advancing the field by open-sourcing all code files and the HiSciVQA dataset for the research community.

Arnav Goel, Medha Hira, Anubha Gupta

Advent of modern deep learning techniques has given rise to advancements in the field of Speech Emotion Recognition (SER). However, most systems prevalent in the field fail to generalize to speakers not seen during training. This study focuses on handling challenges of multilingual SER, specifically on unseen speakers. We introduce CAMuLeNet, a novel architecture leveraging co-attention based fusion and multitask learning to address this problem. Additionally, we benchmark pretrained encoders of Whisper, HuBERT, Wav2Vec2.0, and WavLM using 10-fold leave-speaker-out cross-validation on five existing multilingual benchmark datasets: IEMOCAP, RAVDESS, CREMA-D, EmoDB and CaFE and, release a novel dataset for SER on the Hindi language (BhavVani). CAMuLeNet shows an average improvement of approximately 8% over all benchmarks on unseen speakers determined by our cross-validation strategy.