AINeutralarXiv – CS AI · Jun 26/10
🧠Researchers discovered that large language models fail to refuse harmful requests in low-resource languages not because they lack the underlying safety representations, but because they cannot properly calibrate their safety decisions across languages. A recalibration approach using minimal target-language examples substantially improves refusal rates, suggesting safety alignment failures stem from decision calibration rather than representation gaps.
🧠 Llama
AINeutralarXiv – CS AI · Jun 26/10
🧠Researchers analyzed how large language models process multiple languages through structural representation rather than token-level analysis. The study reveals that low-resource languages have fundamentally different structural properties compared to high-resource languages like English, and that language-specific training alters these structures while maintaining inter-language relationships.
AINeutralarXiv – CS AI · Jun 26/10
🧠Researchers introduce MIDI, a multilingual idiom dataset covering 18 languages across resource tiers, revealing that state-of-the-art NLP models struggle significantly with idiomatic expressions—particularly in low-resource languages and when interpreting literal meanings. The findings expose fundamental gaps in how current AI systems handle contextual language nuance across different linguistic communities.
AINeutralarXiv – CS AI · Jun 26/10
🧠Researchers introduce Luar, a reinforcement learning framework that trains reasoning language models to selectively translate non-English inputs to English only when necessary for reliable reasoning. The approach achieves superior multilingual reasoning performance compared to standard baselines, particularly benefiting low-resource languages while avoiding unnecessary translation overhead.
AINeutralarXiv – CS AI · Jun 15/10
🧠Researchers introduce BEA-Dialogue+, an expanded Hungarian conversational speech recognition corpus that nearly triples training data from 85 to 200 hours while maintaining speaker separation across dataset splits. The expanded resource enables better evaluation of automatic speech recognition models and demonstrates that specialized fine-tuning techniques improve performance on dialogue transcription tasks.
AINeutralarXiv – CS AI · May 295/10
🧠Researchers evaluated nine automatic speech recognition (ASR) models on Dutch child speech datasets, finding that fine-tuned Whisper-medium achieved 5.54% word error rate on clean data but 70.37% on noisy data. Using an utterance-level selection method, they identified 42% of clean recordings as reliable without manual verification, achieving 98.3% precision and significantly reducing annotation overhead for child speech research.
AINeutralarXiv – CS AI · May 286/10
🧠Researchers develop strategies for extending large language models as evaluation tools to multilingual settings, addressing challenges in low-resource languages. The study reveals that fine-tuned smaller models match proprietary performance when in-domain data exists, while larger zero-shot models excel in out-of-domain scenarios, providing practical guidance for building multilingual evaluation systems.
AIBullisharXiv – CS AI · May 276/10
🧠Researchers have released ParsVoice, a 2,200-hour Persian speech dataset with 1.36 million aligned segments from 1,815 speakers, making it 25 times larger than previous Persian TTS resources. The dataset was constructed using an automated pipeline combining ASR, fine-tuned language models, and quality assessment, and validation shows the corpus enables multi-speaker text-to-speech systems competitive with existing solutions.
🏢 Hugging Face
AINeutralarXiv – CS AI · May 126/10
🧠Researchers have developed Bangla-WhisperDiar, a fine-tuned speech recognition and speaker diarization system that achieves a 24.41% word error rate for ASR and 23.92% diarization error rate. The work addresses critical gaps in Bangla language processing by combining OpenAI's Whisper model with PyAnnote's diarization framework, trained on custom datasets with extensive data augmentation techniques.
AINeutralarXiv – CS AI · May 116/10
🧠Researchers propose Multilingual Self-Distillation (MSD), a framework that transfers safety safeguards from high-resource languages like English to vulnerable low-resource languages in large language models. The method eliminates the need for expensive multilingual response data by leveraging an LLM's existing safety capabilities, demonstrating effective cross-lingual protection across diverse jailbreak benchmarks.
AIBullisharXiv – CS AI · May 96/10
🧠Researchers introduced ANGOFA, four pre-trained language models tailored for Angolan languages using Multilingual Adaptive Fine-tuning (MAFT) with OFA embedding initialization and synthetic data. The approach achieved 12.3 and 3.8 point improvements over previous state-of-the-art models, addressing a critical gap in NLP support for very-low resource African languages.
AINeutralarXiv – CS AI · Apr 146/10
🧠Researchers introduce a Cross-Lingual Mapping Task during LLM pre-training to improve multilingual performance across languages with varying data availability. The method achieves significant improvements in machine translation, cross-lingual question answering, and multilingual understanding without requiring extensive parallel data.
AINeutralarXiv – CS AI · Apr 146/10
🧠Researchers identify that reasoning language models exhibit worse performance in low-resource languages due to failures in language understanding rather than reasoning capability itself. The study proposes Selective Translation, which strategically adds English translations only when understanding failures are detected, achieving near full-translation performance while translating just 20% of inputs.
AINeutralarXiv – CS AI · Apr 136/10
🧠Researchers have developed RandSymKL, a debiasing technique for Bangla language models that mitigates gender bias in classification tasks like sentiment analysis and hate speech detection. The study introduces four manually annotated benchmark datasets with gender-perturbation testing and demonstrates that the approach effectively reduces bias while maintaining competitive accuracy compared to existing methods.
AINeutralarXiv – CS AI · Apr 106/10
🧠Researchers evaluated how well large language models can perform formal grammar-based translation tasks using in-context learning, finding that LLM translation accuracy degrades significantly with grammar complexity and sentence length. The study identifies specific failure modes including vocabulary hallucination and untranslated source words, revealing fundamental limitations in LLMs' ability to apply formal grammatical rules to translation tasks.
AIBullisharXiv – CS AI · Apr 76/10
🧠Researchers discovered that multilingual MoE AI models exhibit 'Language Routing Isolation,' where high and low-resource languages activate different expert sets. They developed RISE, a framework that exploits this isolation to improve low-resource language performance by up to 10.85% F1 score while preserving other language capabilities.
AIBullisharXiv – CS AI · Mar 276/10
🧠Researchers successfully fine-tuned LLaMA 3.1-8B for medical transcription in Finnish, a low-resource language, achieving strong semantic similarity despite low n-gram overlap. The study used simulated clinical conversations from students and demonstrates the feasibility of privacy-oriented domain-specific language models for clinical documentation in underrepresented languages.
AIBearisharXiv – CS AI · Mar 36/104
🧠A comprehensive study of 17 Large Language Models as automated annotators for Bangla hate speech detection reveals significant bias and instability issues. The research found that larger models don't necessarily perform better than smaller, task-specific ones, raising concerns about LLM reliability for sensitive annotation tasks in low-resource languages.
AIBullisharXiv – CS AI · Feb 276/107
🧠Researchers developed a new AI framework using RNN-T architecture to improve speech recognition for Taiwanese Hakka, an endangered low-resource language with high dialectal variability. The system achieved 57% and 40% relative error rate reductions for two different writing systems, marking the first systematic investigation into Hakka dialect variations in ASR.
AIBullisharXiv – CS AI · Feb 276/106
🧠Researchers introduced ViCLIP-OT, the first foundation vision-language model specifically designed for Vietnamese image-text retrieval. The model integrates CLIP-style contrastive learning with Similarity-Graph Regularized Optimal Transport (SIGROT) loss, achieving significant improvements over existing baselines with 67.34% average Recall@K on UIT-OpenViIC benchmark.
AIBullisharXiv – CS AI · Feb 275/103
🧠Researchers developed Lipi-Ghor-882, an 882-hour Bengali speech dataset, and demonstrated that targeted fine-tuning with synthetic acoustic degradation significantly improves automatic speech recognition for long-form Bengali audio. Their dual pipeline achieved a 0.019 Real-Time Factor, establishing new benchmarks for low-resource speech processing.
AIBullishMicrosoft Research Blog · Feb 56/103
🧠Microsoft Research launched Paza, a human-centered speech recognition pipeline, and PazaBench, the first benchmark leaderboard specifically designed for low-resource languages. The initiative covers 39 African languages with 52 models and has been tested with real communities to improve AI accessibility for underrepresented languages.
AINeutralarXiv – CS AI · Mar 264/10
🧠Researchers developed Konkani LLM, a specialized language model for the low-resource Indian language Konkani, using a synthetic 100k instruction dataset. The model addresses training data scarcity across multiple scripts (Devanagari, Romi, Kannada) and demonstrates competitive performance against proprietary models in machine translation tasks.
🧠 Gemini🧠 Llama
AINeutralarXiv – CS AI · Mar 54/10
🧠Researchers have developed LilMoo, a 0.6-billion parameter Hindi language model trained from scratch using a transparent, reproducible pipeline optimized for limited compute environments. The model outperforms similarly sized multilingual baselines like Qwen2.5-0.5B and Qwen3-0.6B, demonstrating that language-specific pretraining can rival larger multilingual models.
AIBullisharXiv – CS AI · Mar 44/102
🧠Researchers developed a multistage AI approach for Bengali speech transcription and speaker diarization, achieving significant improvements in processing long-form audio recordings. The system used fine-tuned Whisper models and custom segmentation techniques to address the low-resource nature of Bengali in speech technology applications.
