paper-with-me

홈 › Papers

Raising Bars, Not Parameters: LilMoo Compact Language Model for Hindi

2026-03-03 · Shiza Fatimah, Aniket Sen, Sophia Falk, Florian Mai, Lucie Flek, Nicholas Kluge Corrêa arxiv

The dominance of large multilingual foundation models has widened linguistic inequalities in Natural Language Processing (NLP), often leaving low-resource languages underrepresented. This paper introduces LilMoo, a 0.6-billion-parameter Hindi language model trained entirely from scratch to address this gap. Unlike prior Hindi models that rely on continual pretraining from opaque multilingual foundations, LilMoo is developed through a fully transparent and reproducible pipeline optimized for limited compute environments. We construct a high-quality Hindi corpus (GigaLekh) filtered through both heuristic and learned (LLM-as-a-judge) methods, complemented by bilingual augmentation with curated English data. Using this dataset, we explore various training recipes for small-scale language models. Across comprehensive evaluation suites, LilMoo consistently outperforms comparably sized multilingual baselines such as Qwen2.5-0.5B and Qwen3-0.6B, demonstrating that well-designed language-specific pretraining can rival large multilingual models at the sub-billion-parameter range.

📄 PDF Abstract BibTeX arXiv:2603.03508

Code (0)

등록된 구현이 없습니다.

Tasks

Continual Pretraining

Similar Papers 제목 키워드 기반

Low-power Spike-based Wearable Analytics on RRAM Crossbars

2025-02-10 · Abhiroop Bhattacharjee, Jinquan Shi, Wei-Chen Chen, Xinxin Wang 외

This work introduces a spike-based wearable analytics system utilizing Spiking Neural Networks (SNNs) deployed on an In-memory Computing engine based on RRAM crossbars, which are known for their compactness and energy-ef…

Activity RecognitionHuman Activity Recognition

Optimizing Binary and Ternary Neural Network Inference on RRAM Crossbars using CIM-Explorer

2025-05-20 · Rebecca Pelke, José Cubero-Cascante, Nils Bosbach, Niklas Degener 외

Using Resistive Random Access Memory (RRAM) crossbars in Computing-in-Memory (CIM) architectures offers a promising solution to overcome the von Neumann bottleneck. Due to non-idealities like cell variability, RRAM cross…

Quantization

LoR-LUT: Learning Compact 3D Lookup Tables via Low-Rank Residuals

2026-02-26 · Ziqi Zhao, Abhijit Mishra, Shounak Roychowdhury arxiv

We present LoR-LUT, a unified low-rank formulation for compact and interpretable 3D lookup table (LUT) generation. Unlike conventional 3D-LUT-based techniques that rely on fusion of basis LUTs, which are usually dense te…

Image EnhancementStyle Transfer

Wideband Bandpass Filters Using a Novel Thick Metallization Technology

2019-10-15

A new class of wideband bandpass filters based on using thick metallic bars as microwave resonators, instead of common microstrip lines, is presented. These bars provide a series of advantages over fully planar printed t…

Frequency and Bandwidth Design of FR3-Band Acoustic Filters

2025-05-23 · Taran Anusorn, Omar Barrera, Jack Kramer, Ian Anderson 외

This article presents an approach to control the operating frequency and fractional bandwidth (FBW) of miniature acoustic filters in thin-film lithium niobate (TFLN). More specifically, we used the first-order antisymmet…