Manish Gupta

Principal Applied Researcher at Microsoft. Adjunct Faculty at IIIT Hyderabad. Visiting faculty at ISB Hyderabad. Interested in deep learning, NLP, web mining.

[Manish Gupta @blitzprecision]

InternVL3.5https://youtu.be/Z3mlW_UKShA
Pix2Structhttps://youtu.be/yMYLvblE4EM
VL-Rethinkerhttps://youtu.be/kSl3Y0xaNLs
300+ such videos: https://t.co/pNblVnNDKf

[Manish Gupta @blitzprecision]

𝐃𝐨𝐜𝐐𝐀𝐂: Adaptive Trie-Guided Decoding for Effective In-Document Query Auto-Completion -- https://t.co/tkH3gJNJwu
300+ such videos: https://t.co/pNblVnNDKf

[Manish Gupta @blitzprecision]

𝐌𝐮𝐥𝐭𝐢-𝐃𝐞𝐜𝐨𝐝𝐞𝐫 𝐅𝐢𝐧𝐞-𝐓𝐮𝐧𝐢𝐧𝐠 𝐟𝐨𝐫 𝐌𝐮𝐥𝐭𝐢-𝐇𝐨𝐩 𝐕𝐐𝐆 𝐰𝐢𝐭𝐡 𝐄𝐱𝐭𝐞𝐫𝐧𝐚𝐥 𝐊𝐧𝐨𝐰𝐥𝐞𝐝𝐠𝐞 https://t.co/G1CyG1eS8w
𝐋𝐨𝐑𝐀 𝐨𝐧 𝐭𝐡𝐞 𝐆𝐨: Instance-level Dynamic LoRA Selection and Merging https://t.co/iaUfopri1M
https://t.co/pNblVnNDKf

[Manish Gupta @blitzprecision]

𝐂𝐡𝐚𝐢𝐧 𝐨𝐟 𝐂𝐨𝐧𝐭𝐢𝐧𝐮𝐨𝐮𝐬 𝐓𝐡𝐨𝐮𝐠𝐡𝐭 -- https://t.co/yKINnfLOSa
𝐓𝐞𝐱𝐭𝟐𝐀𝐫𝐜𝐡: Generating Architecture Diagrams from Natural Language Descriptions -- https://t.co/1a115VCUBh
300+ such videos: https://t.co/pNblVnNDKf

[Manish Gupta @blitzprecision]

DoRA: Weight-Decomposed Low-Rank Adaptation -- https://t.co/B36VxzcJiS
Test-Time Training -- https://t.co/BzvU1HwG08
DeepSeekMath and RL for LLMs -- https://t.co/lWI96o7tLP
300+ such videos: https://t.co/pNblVnNDKf

[Manish Gupta @blitzprecision]

𝐉𝐨𝐢𝐧𝐭-𝐄𝐦𝐛𝐞𝐝𝐝𝐢𝐧𝐠 𝐏𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐯𝐞 𝐀𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞 (𝐉𝐄𝐏𝐀) -- https://t.co/9YcLtGxkAo
𝐀𝐋𝐢𝐁𝐢 (Attention with Linear Biases) -- https://t.co/sYstW1KLeC
300+ such videos: https://t.co/pNblVnNDKf

[Manish Gupta @blitzprecision]

𝐀𝐦𝐚𝐳𝐨𝐧 𝐂𝐡𝐫𝐨𝐧𝐨𝐬: LLMs for time series https://t.co/tjwNiJGA5u
𝐉𝐨𝐢𝐧𝐭-𝐄𝐦𝐛𝐞𝐝𝐝𝐢𝐧𝐠 𝐏𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐯𝐞 𝐀𝐫𝐜𝐡𝐢𝐭𝐞𝐜𝐭𝐮𝐫𝐞 (𝐉𝐄𝐏𝐀) https://t.co/9YcLtGxkAo

300+ videos: https://t.co/pNblVnNDKf

[Manish Gupta @blitzprecision]

𝐅𝐚𝐫𝐚-𝟕𝐁 Computer Use Agent -- https://t.co/ogfFiKeLLR
𝐓𝐢𝐭𝐚𝐧𝐬: Learning to Memorize at Test Time -- https://t.co/5CUTEgjyH2
𝐁𝐞𝐫𝐤𝐞𝐥𝐞𝐲 𝐅𝐮𝐧𝐜𝐭𝐢𝐨𝐧 𝐂𝐚𝐥𝐥𝐢𝐧𝐠 𝐋𝐞𝐚𝐝𝐞𝐫𝐛𝐨𝐚𝐫𝐝 -- https://t.co/8Ra87A8MHg
250+ videos: https://t.co/pNblVnNDKf

[Manish Gupta @blitzprecision]

𝐋𝐚𝐫𝐠𝐞 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐃𝐢𝐟𝐟𝐮𝐬𝐢𝐨𝐧 𝐌𝐨𝐝𝐞𝐥𝐬 -- https://t.co/BK66ekrMZp
𝐇𝐢𝐯𝐞𝐦𝐢𝐧𝐝 𝐄𝐟𝐟𝐞𝐜𝐭 𝐢𝐧 𝐋𝐋𝐌𝐬 -- https://t.co/iFSc24udEp
𝐆𝐚𝐭𝐞𝐝 𝐀𝐭𝐭𝐞𝐧𝐭𝐢𝐨𝐧 𝐟𝐨𝐫 𝐋𝐋𝐌𝐬 -- https://t.co/CB4FZhm4NV
250+ such videos at https://t.co/pNblVnNDKf

[Manish Gupta @blitzprecision]

𝐂𝐨𝐑𝐓: Teaching Language Models to Reason with Tools -- https://t.co/bxTwpDRV4H
𝐨𝐥𝐦𝐎𝐂𝐑-𝟐 -- https://t.co/qH8PO8oWet
𝐇𝐀𝐋𝐨𝐆𝐄𝐍: LLM Hallucinations -- https://t.co/BxDORbBI9N
250+ such videos: https://t.co/pNblVnNDKf

[Manish Gupta @blitzprecision]

FRAMES dataset -- https://t.co/vcQl2AjhEZ
Attention Sinks -- https://t.co/84q9gioKSE
250+ such videos: https://t.co/pNblVnN5UH

[Manish Gupta @blitzprecision]

DeepSeek-OCR -- https://t.co/wN0n5jeQj0
Humanity's Last Exam -- https://t.co/eHyfGwnXOe
BrowseComp -- https://t.co/cw4NSaDQxx
250+ such videos: https://t.co/pNblVnN5UH

[Manish Gupta @blitzprecision]

𝐀𝐝𝐚𝐩𝐭𝐢𝐯𝐞 𝐊𝐕 𝐂𝐚𝐜𝐡𝐞 𝐂𝐨𝐦𝐩𝐫𝐞𝐬𝐬𝐢𝐨𝐧 for LLMs -- https://t.co/jwkrIzfUN7
𝐍𝐚𝐭𝐢𝐯𝐞 𝐒𝐩𝐚𝐫𝐬𝐞 𝐀𝐭𝐭𝐞𝐧𝐭𝐢𝐨𝐧 -- https://t.co/5el99rlKYn
𝐊𝐢𝐝𝐋𝐌: LLMs for Children -- https://t.co/Y0fTkjWEou
250+ such videos: https://t.co/pNblVnN5UH

[Manish Gupta @blitzprecision]

𝐑𝐞𝐬𝐭𝐨𝐫𝐦𝐞𝐫: Image Restoration -- https://t.co/3KKe33lalf
𝐒𝐜𝐚𝐥𝐢𝐧𝐠 𝐋𝐚𝐰𝐬 𝐅𝐨𝐫 𝐃𝐞𝐧𝐬𝐞 𝐑𝐞𝐭𝐫𝐢𝐞𝐯𝐚𝐥 -- https://t.co/ADplm36RrR
𝐑𝐞𝐬𝐩𝐨𝐧𝐬𝐞 𝐒𝐚𝐦𝐩𝐥𝐢𝐧𝐠 𝐢𝐧 𝐋𝐋𝐌𝐬 -- https://t.co/A8ALlgLT4t
250+ such videos: https://t.co/pNblVnNDKf

[Manish Gupta @blitzprecision]

𝐊𝐈𝐌𝐈 𝐊𝟐: Open Agentic Intelligence -- https://t.co/kCWLfgw2rj
𝐃𝐢𝐧𝐨𝐕𝟑 -- https://t.co/ShssN7mLaX
𝐌𝐞𝐭𝐚'𝐬 𝐑𝐞𝐅𝐫𝐚𝐠: Compress, Sense and Expand for 31x faster Decoding -- https://t.co/avVZHCMpQx
250+ such videos: https://t.co/pNblVnNDKf

[Manish Gupta @blitzprecision]

Paper accepted: @anujanegi_1 , @subbareddy300 , @anwarnunez, @blitzprecision, Fatma Deniz. Brain-Informed Fine-Tuning for Improved Multilingual Understanding in Language Models.  NeurIPS 2025. https://t.co/DLPyxS8wgM

[Manish Gupta @blitzprecision]

𝐓𝐞𝐱𝐭 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐨𝐧 𝐰𝐢𝐭𝐡 𝐃𝐢𝐟𝐟𝐮𝐬𝐢𝐨𝐧 𝐋𝐚𝐧𝐠𝐮𝐚𝐠𝐞 𝐌𝐨𝐝𝐞𝐥𝐬 -- https://t.co/YinlrXRy1R
𝐕𝐢𝐬𝐢𝐨𝐧 𝐓𝐫𝐚𝐧𝐬𝐟𝐨𝐫𝐦𝐞𝐫𝐬 𝐰𝐢𝐭𝐡 𝐑𝐞𝐠𝐢𝐬𝐭𝐞𝐫𝐬 -- https://t.co/BRsSrt3pzn
250+ videos at https://t.co/pNblVnN5UH

[Manish Gupta @blitzprecision]

𝐆𝐏𝐓𝟓 -- https://t.co/cRoaUP5vRB
𝐆𝐞𝐦𝐦𝐚 𝟑 -- https://t.co/vjwOHDoLNw
250+ videos at https://t.co/pNblVnN5UH

[Manish Gupta @blitzprecision]

B. Gaur, K. Gupta, A. Srivastava, M. Gupta, Md S. Akhtar. 𝐀𝐬𝐬𝐞𝐬𝐬 𝐚𝐧𝐝 𝐏𝐫𝐨𝐦𝐩𝐭: 𝐀 𝐆𝐞𝐧𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐑𝐋 𝐅𝐫𝐚𝐦𝐞𝐰𝐨𝐫𝐤 𝐟𝐨𝐫 𝐈𝐦𝐩𝐫𝐨𝐯𝐢𝐧𝐠 𝐄𝐧𝐠𝐚𝐠𝐞𝐦𝐞𝐧𝐭 𝐢𝐧 𝐎𝐧𝐥𝐢𝐧𝐞 𝐌𝐞𝐧𝐭𝐚𝐥 𝐇𝐞𝐚𝐥𝐭𝐡 𝐂𝐨𝐦𝐦𝐮𝐧𝐢𝐭𝐢𝐞𝐬.
https://t.co/B2knhJkT3d

[Manish Gupta @blitzprecision]

Got the following 4 papers accepted at EMNLP 2025

[Manish Gupta @blitzprecision]

T Karthikeyan, O. Dehlan, Mausam, M. Gupta. 𝐋𝐑𝐏𝐋𝐀𝐍: 𝐀 𝐌𝐮𝐥𝐭𝐢-𝐀𝐠𝐞𝐧𝐭 𝐂𝐨𝐥𝐥𝐚𝐛𝐨𝐫𝐚𝐭𝐢𝐨𝐧 𝐨𝐟 LLMs 𝐚𝐧𝐝 LRMs 𝐟𝐨𝐫 𝐏𝐥𝐚𝐧𝐧𝐢𝐧𝐠 𝐰𝐢𝐭𝐡 𝐈𝐦𝐩𝐥𝐢𝐜𝐢𝐭 & 𝐄𝐱𝐩𝐥𝐢𝐜𝐢𝐭 𝐂𝐨𝐧𝐬𝐭𝐫𝐚𝐢𝐧𝐭𝐬.
https://t.co/jbOTcZvbZK