CognitiveLab
@cognitivelab_ai
Followers
1K
Following
24
Media
8
Statuses
24
Democratizing Generative AI
India
Joined July 2023
Read more : https://t.co/t5pEav2h92 We thank @modal for sponsoring compute for this research.
cognitivelab.in
We're excited to announce NetraEmbed and ColNetraEmbed, achieving 152% improvement over existing systems in multilingual document retrieval. Supporting 22 languages with state-of-the-art performance.
0
2
16
Also excited to release "M3DR: Towards Universal Multilingual Multimodal Document Retrieval" paper preprint that cover the framework we used to train these models
1
3
23
Achieves SoTA performance on both cross lingual and monolingual benchmarks on our newly released NayanaIR-Bench
1
1
16
Today we are excited to launch NetraEmbed SoTA multimodal multilingual document retrieval model. > Supports 22 languages > ~ 150% improvement over existing baselines > NayanaIR-Bench a open source multilingual document retrieval benchmark
5
28
193
Happy to share got a paper accepted at a neurips workshop This paper mainly tackles long context reasoning for multimodal models(vlms) with key focus citations and reasoning based re ranking got some good results with sft + rl(grpo) will put it on arxiv soon!
23
6
294
Nayanaโs Mission: Democratize AI for education, healthcare, governance, & cultural preservation across languages & modalities. Official Announcement : https://t.co/qRloKAZvhj
about.fb.com
Weโre excited to introduce the 10 international recipients of the second Llama Impact Grants.
1
0
7
@cognitivelab_ai latest Milestones: ๐ Poster at #llamacon ๐ Papers accepted at NAACL & CVPR workshops ๐ Training on millions of synthetic datasets We're replacing fragmented AI pipelines with ONE cohesive model!
1
0
4
Meet Nayana ("eyes" in Sanskrit): A unified AI model for text, vision, & audio! โ
Supports 22 languages (10 Indic + 12 global) โ
OCR, translation, Q&A, summarization many more tasks Know more :
cognitivelab.in
CognitiveLab is proud to be selected as a recipient of Meta's prestigious Llama Impact Grant 2024, accelerating our revolutionary Nayana projectโa multilingual (22 languages, 10 Indic), multimodal AI...
1
0
6
๐ Thrilled to announce CognitiveLab wins the six-figure Llama Impact Grant by @Meta ! ๐ฎ๐ณ As India's only recipient, we're powering Nayana a multimodal, multilingual multi task AI model family! #AI #MultilingualAI #Innovation
2
0
34
๐๐ผ Finetune for multi turn chat conversation ๐๐ผ Performers decent in cross lingual tasks ๐๐ผ Will soon put out the evals on Indic LLM leader board https://t.co/C2FWveiRro
huggingface.co
0
0
3
Our first attempt at fine-tuning LLama3 for Indic languages. we choose Hindi as LLama3 was efficient at tokenising it Many more experiments/iterations are in progress
1
0
1
๐จ Introducing Gaja ๐จ ~ (Llama3-Gaja) A series of open source bilingual Hindi-English LLMs finetuned on top of Llama3-8b by @AIatMeta
https://t.co/EJnjVgP387
1
8
75
evaluated LLama3 using indic_eval ( https://t.co/KFEXLwzxFM) on English, Hindi and Kannada results are there on the Indic LLM Leader board https://t.co/G8OfxTyCzV ๐๐ผ It performs significantly better than LLama2 in most of the benchmarks. ๐๐ผ In comparison to Gemma, it falls a
huggingface.co
It's going to be hard to adapt Llama3 for Indic languages, in my opinion. Here are a few reasons why: ๐๐ผ The tokenizer used is TikToken-based, which is not really efficient in tokenizing Indic text despite having a vocabular size of 121k. ๐๐ผ unlike sentence-piece based models,
1
3
29
The heart of the Leaderboard is indic_eval - https://t.co/kRcitPXXNo is an evaluation library that integrates seamlessly with the indic LLM leaderboard.
github.com
A lightweight evaluation suite tailored specifically for assessing Indic LLMs across a diverse range of tasks - adithya-s-k/indic_eval
0
0
1
Features include: ๐๐ผ Support for 7 Indic languages. ๐๐ผ Open source, hosted on Hugging Face. ๐๐ผ Support for 4 Indic benchmarks, with more to be added. ๐๐ผ Seamless integration with indic_eval This is the alpha release and we will be adding a lot more tested features soon
1
0
1
Impressive Indic LLMs are emerging weekly, but without a standardized evaluation method or leaderboard, meaningful comparisons are difficult, hampering research and innovation. ๐๐ผ That's why we built - an open-source platform for comparing Indic LLMs across various benchmarks.
1
0
0
Introducing ๐จIndic LLM Leaderboard๐จ (alpha release)
1
5
17
Here is the blog going into details about the model https://t.co/cG7LM6CX9X You can find the models here https://t.co/MCScJ6s0RS
huggingface.co
0
0
7
Our inaugural models, ๐๐บ๐ฏ๐ฎ๐ฟ๐ถ-๐ณ๐-๐ฏ๐ฎ๐๐ฒ-๐๐ฌ.๐ญ and ๐๐บ๐ฏ๐ฎ๐ฟ๐ถ-๐ณ๐-๐๐ป๐๐๐ฟ๐๐ฐ๐-๐๐ฌ.๐ญ, achieve impressive results on a compact 1 billion-token training dataset, trained across multiple stages.
1
0
5