// BLOG

Research & updates

Exploring the frontiers of AI, language models, and multilingual technology

1st May, 202601

When LLMs Stop Following Steps: A Diagnostic Study of Procedural Execution in Language Models

EACL 2027 Main Conference. A diagnostic study of how well language models execute step-by-step procedures.

Read more
3rd February, 202602

Task–Specificity Score: Measuring How Much Instructions Really Matter for Supervision

EMNLP 2026. Measuring how much instructions really matter for supervision.

Read more
12th November, 202503

CoSHE-Eval: A Code-Switching ASR Benchmark for Hindi–English Speech

CoSHE-Eval is a 30-hour Hindi–English code-switching evaluation dataset designed to benchmark ASR systems under realistic multilingual speech conditions.

Read more
6th November, 202504

Dhrith: Emotionally Intelligent ASR for India's Multilingual Voices

Dhrith is our next-generation ASR model that listens beyond words. It understands emotion, rhythm, and code-switched language.

Read more
Introducing Pragna-1B: Soket AI Labs' Multilingual Language Model for Indian Languages
30th April, 202405

Introducing Pragna-1B: Soket AI Labs' Multilingual Language Model for Indian Languages

We at Soket AI Labs are thrilled to unveil India's first open source multilingual model, Pragna-1B available in four Indian languages.

Read more
Availability of the Bhasha SFT Dataset for Supervised Fine-Tuning of Indic Language Models
19th April, 202406

Availability of the Bhasha SFT Dataset for Supervised Fine-Tuning of Indic Language Models

An extensive collection curated by Soket AI Labs for the supervised fine-tuning of Multilingual Large Language Models, focusing on Indic languages.

Read more
Introducing the "Bhasha" Series: Advancements in Indic Language AI Datasets
17th April, 202407

Introducing the "Bhasha" Series: Advancements in Indic Language AI Datasets

The "Bhasha" series datasets are engineered to support the development of AI models attuned to the linguistic and cultural nuances of India.

Read more