low-resource

This paper describes a two-stage vision-language adaptation system for Nepali meme classification, using Qwen3-VL-8B-Instruct with LoRA fine-tuning and contrastive learning. The system achieved 2nd place in hate speech detection and 4th in sentiment analysis at the CHiPSAL 2026 shared task.

This paper evaluates forced alignment for Hindi-English code-mixed speech using the Montreal Forced Aligner, demonstrating that bootstrapping strategies and code-mixed training data achieve a tenfold improvement in alignment accuracy over monolingual alternatives.

Aletheia is an offline-first clinical decision support system fine-tuned from Qwen2.5-3B-Instruct using QLoRA on 27,000 clinical reasoning samples for low-resource healthcare settings in sub-Saharan Africa, achieving 80% Top-1 accuracy and fitting within memory constraints.

This paper compares fine-tuned MahaBERT-based models with large language models (Gemini, LLaMA-3.3-70B, Gemma) for Marathi named entity recognition, finding that the specialized BERT models significantly outperform the LLMs, achieving F1-scores of 0.88–0.91 versus 0.57–0.69.

This paper presents a systematic study of cross-lingual transfer for biomedical machine translation into low-resource Arabic-script languages, using Arabic and Persian as pivots. The authors evaluate LoRA adapter merging as a zero-data transfer strategy, showing it works surprisingly well for closely related languages like Dari.

Introduces Khondo, the first benchmark for document packet splitting on Bangladeshi government forms. It is a vision-native, bilingual dataset spanning multiple concatenation schemes, and zero-shot evaluation of MLLMs reveals that page-order reconstruction remains a key open challenge.

This paper proposes a cross-lingual data augmentation strategy that uses machine translation to transfer expert-annotated difficulty labels from high-resource languages to low-resource languages. Experiments with BERT-based regression models show that augmenting scarce native data with translated corpora significantly improves the accuracy of text difficulty assessment.

This paper proposes a rationale-guided knowledge distillation framework for cross-lingual stance detection, using chain-of-thought prompting from large language models to train a compact student model with dual-path distillation and contrastive learning.

LatentMT applies latent-reasoning looped language models to machine translation, achieving performance comparable to models three to five times larger while requiring lower compute, and sets state-of-the-art on mid- and low-resource languages.

FedCC proposes a federated learning framework combining a frozen DINOv2 backbone with a lightweight YOLO detection head and LoRA modules for robust corpus callosum localization in fetal ultrasound images, achieving strong performance with greatly reduced communication cost in a multi-center setting.

This paper introduces a benchmark dataset of 1,516 expert-verified Bangla sentences for disambiguating culturally entangled homographs (words that are both names and common nouns). It shows that LLMs suffer from dominant-meaning bias and proposes contrastive chain-of-thought prompting and distillation to reduce this bias.

This paper investigates cross-lingual clinical correctness drift in medical language models, finding that locally deployable models show significant safety degradation when queried in Hausa compared to English, while frontier models maintain competence, highlighting a critical gap in safety evaluation for low-resource settings.

This paper introduces BLAD, a curated multilingual dataset of 1,484 Bangladeshi legal acts spanning 1799 to 2025, with structured metadata for temporal and cross-lingual legal NLP research.

This paper proposes two hybrid continual learning methods—Replay Augmented Elastic Weight Consolidation and Constraint Guided Knowledge Distillation—to adapt pretrained speech models for identifying low-resource Australian Aboriginal languages while mitigating catastrophic forgetting.

This paper presents a hybrid approach for detecting online polarization in English and Hausa using DeBERTa for English and AfroXLMR-Social for Hausa and fine-grained subtasks, with LoRA and data augmentation to address computational and data constraints.

This paper investigates cross-lingual transfer for low-resource ASR in Warlpiri, proposing a similarity-based framework combining acoustic and linguistic features to select optimal source languages. Experiments show that acoustically similar languages like Assamese and Hindi significantly reduce word and character error rates.

This paper presents a sentence-level sign language translation system fine-tuned with QLoRA on a subset of How2Sign, achieving BLEU 15.9. Its main contribution is a hardware-aware streaming pipeline using a Raspberry Pi 4B client and a CPU/GPU backend, reducing mean latency by 27.71%.

This paper proposes a multi-conditioned diffusion-based synthesis pipeline using Stable Diffusion XL and ControlNet to generate synthetic sand boil imagery for low-resource earthen-levee inspection, addressing the scarcity of annotated defect examples.

Colibri runs the 744B parameter GLM-5.2 MoE model on a laptop with 25GB RAM by activating only ~40B parameters per token and streaming experts from disk, all in a single 2,400-line C file with no GPU required.

This paper introduces the Nigeria Machinery Usage and Failures Dataset, 89 records across 28 indicators for Nigeria's manufacturing and oil/gas sectors from 2006 to 2025, along with a method to build domain-grounded chain-of-thought reasoning examples from sparse numeric values.