Korean
PulseAugur coverage of Korean — every cluster mentioning Korean across labs, papers, and developer communities, ranked by signal.
9 day(s) with sentiment data
-
NVIDIA Magpie TTS adds 3 languages, now supports 12 total
NVIDIA has expanded its Magpie Text-to-Speech (TTS) model to support 12 languages, including the recent additions of Arabic, Korean, and Brazilian Portuguese. This open-weights model, featuring 364 million parameters, i…
-
AI Community Innovates with MiniMax Distillation, New ASR Model, and Gemini Rumors
The AI community has rapidly developed a distilled LoRA model based on MiniMax AI's recently released weights, showcasing rapid innovation. Separately, a new open-source model, Audio8-ASR-0.1B, has been released, offeri…
-
New distillation technique boosts multilingual math reasoning in LLMs
Researchers have explored On-Policy Delta Distillation (OPD^2), an advancement over On-Policy Distillation (OPD), for multilingual mathematical reasoning. Experiments using the Qwen3 model demonstrated that OPD^2 signif…
-
LLM subtitle translation workflow requires multi-stage validation beyond good prompts
Building a production-grade multilingual subtitle translation workflow involves more than just crafting a good prompt. The process requires a multi-stage approach, including initial translation, deterministic validation…
-
New NOLLI benchmark reveals English-Korean LLM performance gaps
Researchers have developed NOLLI, a new benchmark designed to pinpoint performance discrepancies between English and Korean language models. The benchmark features 15 puzzle types and 7,500 items, with difficulty calibr…
-
LLM essays outperform student essays on Korean writing rubric
A recent study published on arXiv explored the differences between essays written by students and those generated by large language models (LLMs) using a Korean writing rubric. The research found that LLM-generated essa…
-
New TIDES dataset enhances AI's understanding of multi-party conversations
Researchers have introduced TIDES, a new longitudinal dataset designed to improve the modeling of multi-party social dynamics in group conversations. This dataset comprises over 75,000 utterances in English and Korean f…
-
Open-source bamboo cooler design inspired by Korean devices released
An open-source design for a human cooler, inspired by traditional Korean devices, has been released. This cooler features an artisan-made bamboo lattice and is presented as a sustainable option for personal cooling. Pla…
-
PDF parsing quality is the hidden bottleneck for RAG systems
The quality of PDF parsing significantly impacts the performance of Retrieval-Augmented Generation (RAG) systems, often acting as a hidden bottleneck. Poor parsing can corrupt text chunks with encoding errors or flatten…
-
New Korean instruction corpus released, generated by Phi-3.5-MoE-instruct
Researchers have released GLAN-QnA-KR, a large-scale Korean instruction-QA corpus containing over 300,000 rows. This corpus was generated using Microsoft's Phi-3.5-MoE-instruct model and a seedless taxonomy-driven synth…
-
New AEGIS framework explores span-guided multilingual text detoxification
A new framework called AEGIS has been developed to study the effectiveness of span-level guidance in multilingual text detoxification. This framework combines span detector outputs with frozen generator backbones, allow…
-
New ROK-FORTRESS benchmark measures LLM safety across languages and geopolitics
A new benchmark called ROK-FORTRESS has been developed to evaluate the safety of large language models (LLMs) in high-stakes national security and public safety contexts, specifically focusing on the interaction between…
-
CoPiT pipeline boosts low-resource Mongolian translation accuracy
Researchers have developed CoPiT, a novel translation pipeline designed to address the challenges of low-resource languages, specifically focusing on Mongolian. This system leverages the imbalance in data availability b…
-
Korean adnominal ending constructions analyzed beyond relative clauses
This paper introduces a new corpus-based typology for annotating Korean adnominal ending constructions, moving beyond simple relative-clause identification. The research argues that the Korean ending 'ETM' is not solely…
-
Solo Leveling movie announced for theatrical release ahead of season 3
Crunchyroll announced at Anime Expo 2026 that a new movie based on the popular Korean comic "Solo Leveling" will be released in theaters. This movie will precede the anime's third season, marking a shift from a typical …
-
Region-specific calibration boosts smartphone dermatology color reliability
A new study published on arXiv details a method for improving the colorimetric reliability of smartphone-based dermatology. Researchers found that region-specific calibration, applied in the CIELAB color space, signific…
-
Cohere and LG CNS unveil LuckyStar 111B for Korean-English agents
Cohere and LG CNS have collaborated to develop LuckyStar 111B, a large language model designed for Korean-English enterprise agents. This model is built upon Cohere's Command A model and employs techniques like preamble…
-
New Korean Negation Benchmark 'Thunder-KoNUBench' Released for LLMs
Researchers have introduced Thunder-KoNUBench, a new benchmark designed to evaluate the negation understanding capabilities of large language models (LLMs) specifically in Korean. The benchmark was developed through a c…
-
Samsung, SK Group plan $1.3T investment in AI, semiconductors
Samsung and SK Group are reportedly planning a significant investment of up to 2 quadrillion Korean won (approximately $1.3 trillion USD) over the next decade. This substantial capital infusion is slated to focus on key…
-
New pipeline KG2Cypher builds natural-language interfaces for enterprise knowledge graphs
Researchers have developed KG2Cypher, a data-centric pipeline designed to create natural-language interfaces for enterprise knowledge graphs. The system generates executable Cypher queries from graph facts and then uses…