𝐅𝐢𝐧𝐝𝐢𝐧𝐠 𝐭𝐡𝐞 𝐫𝐢𝐠𝐡𝐭 𝐦𝐨𝐦𝐞𝐧𝐭 𝐢𝐧 𝐚 𝐯𝐢𝐝𝐞𝐨 𝐮𝐬𝐢𝐧𝐠 𝐚𝐧𝐨𝐭𝐡𝐞𝐫 𝐯𝐢𝐝𝐞𝐨? 𝐒𝐨𝐮𝐧𝐝𝐬 𝐡𝐚𝐫𝐝. 𝐈𝐧𝐝𝐞𝐞𝐝 𝐢𝐭 𝐢𝐬! Paper accepted at ICCV 2025: Aligning Moments in Time using Video Queries. 𝐏𝐫𝐞-𝐩𝐫𝐢𝐧𝐭:

𝐌𝐞𝐚𝐬𝐮𝐫𝐢𝐧𝐠 𝐂𝐮𝐥𝐭𝐮𝐫𝐚𝐥 𝐁𝐢𝐚𝐬 𝐢𝐧 𝐋𝐋𝐌𝐬 -- Happy learning! :-).

𝐇𝐨𝐰 𝐭𝐨 𝐏𝐞𝐫𝐬𝐮𝐚𝐝𝐞 𝐋𝐋𝐌𝐬 𝐭𝐨 𝐉𝐚𝐢𝐥𝐛𝐫𝐞𝐚𝐤 𝐓𝐡𝐞𝐦? -- 𝐒𝐞𝐚𝐫𝐜𝐡-𝐨𝟏: 𝐀𝐠𝐞𝐧𝐭𝐢𝐜 𝐒𝐞𝐚𝐫𝐜𝐡-𝐄𝐧𝐡𝐚𝐧𝐜𝐞𝐝 𝐋𝐚𝐫𝐠𝐞 𝐑𝐞𝐚𝐬𝐨𝐧𝐢𝐧𝐠 𝐌𝐨𝐝𝐞𝐥𝐬 --

𝐂𝐚𝐧 𝐋𝐋𝐌𝐬 𝐥𝐞𝐚𝐫𝐧 𝐬𝐲𝐧𝐭𝐡𝐞𝐭𝐢𝐜 𝐥𝐚𝐧𝐠𝐮𝐚𝐠𝐞𝐬 𝐞𝐪𝐮𝐚𝐥𝐥𝐲 𝐰𝐞𝐥𝐥? -- 𝐀𝐲𝐚 𝐌𝐨𝐝𝐞𝐥-- Happy learning! :-).

Paper accepted at ICML 2025 on "𝐌𝐎𝐆𝐈𝐂: 𝐌𝐞𝐭𝐚𝐝𝐚𝐭𝐚-𝐈𝐧𝐟𝐮𝐬𝐞𝐝 𝐎𝐫𝐚𝐜𝐥𝐞 𝐆𝐮𝐢𝐝𝐚𝐧𝐜𝐞 𝐅𝐨𝐫 𝐈𝐦𝐩𝐫𝐨𝐯𝐞𝐝 𝐄𝐱𝐭𝐫𝐞𝐦𝐞 𝐂𝐥𝐚𝐬𝐬𝐢𝐟𝐢𝐜𝐚𝐭𝐢𝐨𝐧." .

Llama-4 models: Rho-1: Selection Language Modeling: Graph of Thoughts for LLMs: .Visual Autoregressive Modeling for Image Generation: 200+ such videos on

DeepSeek-V3 -- Phi-4 -- Large Language Models are Human-like Annotators -- PRISM Alignment Dataset for LLMs -- Happy learning! :-).

𝐕𝐢𝐝𝐞𝐨𝐩𝐨𝐞𝐭: LLM for zero-shot video generation -- 𝐒𝐭𝐞𝐚𝐥𝐢𝐧𝐠 𝐩𝐚𝐫𝐭 𝐨𝐟 𝐚𝐧 𝐋𝐋𝐌 -- 𝐃𝐞𝐞𝐩𝐒𝐞𝐞𝐤-𝐑𝟏 -- Watch 200+ such videos on

3/3: 2 full paper accepts at ICLR. Links to papers: .

2/3: 2 full paper accepts at ICLR. @subbareddy300 @KhushbuPahwa @MounikaMarreddy @ManeeshKSingh5 @blitzprecision, Bapi Raju. Multi-modal brain encoding models for multi-modal stimuli.

1/3: 2 full paper accepts at ICLR. @subbareddy300, Akshett Jindal, @tweepy_ishani, @khushbu_pahwa, @srinath_namburi, @manshri, @ManeeshKSingh5, Bapi Raju, @blitzprecision. Correlating instruction-tuning (in multimodal models) with vision-language processing (in the brain).

I created a playlist on 𝐌𝐮𝐥𝐭𝐢-𝐯𝐞𝐜𝐭𝐨𝐫 𝐝𝐞𝐧𝐬𝐞 𝐫𝐞𝐭𝐫𝐢𝐞𝐯𝐚𝐥 methods: 𝐂𝐨𝐥𝐁𝐄𝐑𝐓, 𝐗𝐓𝐑, 𝐂𝐨𝐥𝐁𝐄𝐑𝐓𝐯𝟐, 𝐏𝐋𝐀𝐈𝐃, 𝐌𝐔𝐕𝐄𝐑𝐀. Watch these and more on my YouTube with 200+ such videos #denseRetrieval.

𝐆𝐞𝐧𝐢𝐞: Generative Interactive Environments: 𝐒𝐭𝐚𝐛𝐥𝐞 𝐃𝐢𝐟𝐟𝐮𝐬𝐢𝐨𝐧 𝟑 (𝐮𝐬𝐢𝐧𝐠 𝐌𝐌-𝐃𝐢𝐓): 𝐌𝐚𝐭𝐅𝐨𝐫𝐦𝐞𝐫: Nested Transformer for Elastic Inference: More at

Tutorial on AlphaFold detailed architecture:.Part1: Part2: Happy learning! :-).

Awesome panel discussion with Sandipan Dandapat, Shirish Karande, Srinivas Bangalore, Prof. @devvarma on multiple aspects on usage of LLMs across industry: myths/misconceptions, value add, ways of using LLMs for various sectors, and also on NLP besides LLMs at @iiit_hyderabad.

Weekend learning (2/2):.𝐍𝐕𝐄𝐦𝐛𝐞𝐝: Best text embedding model as of now -- 𝐌𝐢𝐧𝐢𝐂𝐏𝐌-𝐕: A GPT-4V Level MLLM on Your Phone -- 𝐍𝐞𝐮𝐫𝐚𝐥 𝐂𝐨𝐫𝐩𝐮𝐬 𝐈𝐧𝐝𝐞𝐱𝐞𝐫 --

Weekend learning:.𝐋𝐥𝐚𝐦𝐚𝟑.𝟐 -- 𝐈𝐦𝐚𝐠𝐞𝐧𝟑 -- 𝐎𝐩𝐞𝐧𝐀𝐈 𝐨𝟏 -- 𝐎𝐏𝐑𝐎: Large Language Models as Optimizers --

𝐖𝐡𝐢𝐥𝐞 𝐥𝐢𝐬𝐭𝐞𝐧𝐢𝐧𝐠 𝐭𝐨 𝐚 𝐬𝐭𝐨𝐫𝐲, 𝐰𝐡𝐢𝐜𝐡 𝐥𝐢𝐧𝐠𝐮𝐢𝐬𝐭𝐢𝐜 𝐩𝐫𝐨𝐩𝐬 𝐨𝐟 𝐄𝐧𝐠𝐥𝐢𝐬𝐡 𝐡𝐞𝐥𝐩 𝐨𝐮𝐫 𝐛𝐫𝐚𝐢𝐧𝐬 𝐮𝐧𝐝𝐞𝐫𝐬𝐭𝐚𝐧𝐝 𝐭𝐡𝐞 𝐬𝐭𝐨𝐫𝐲?Full paper at NeuRIPS with Subba and Mariya.Joint processing of linguistic props in brains and LMs.

How can you create models with 𝟏𝟓𝐱 𝐦𝐨𝐫𝐞 𝐭𝐡𝐫𝐨𝐮𝐠𝐡𝐩𝐮𝐭 𝐜𝐨𝐦𝐩𝐚𝐫𝐞𝐝 𝐭𝐨 𝐋𝐥𝐚𝐦𝐚 𝟐 but can still give similar accuracies? How do you generate audio for text caption like "a dog barks while somebody plays the trumpet in a busy street"?