Soham Deshmukh

116 posts

Soham Deshmukh

Soham Deshmukh

@sohamdesh_

language and audio @Sesame Prev: @Microsoft @CarnegieMellon

Redmond, WA Katılım Temmuz 2015
257 Takip Edilen188 Takipçiler
Soham Deshmukh retweetledi
Satvik Dixit
Satvik Dixit@SatvikDixit9·
Our paper "Mellow: a small audio language model for reasoning" was accepted to #NeurIPS2025! Catch our poster in Session 3 at 11am today.
Soham Deshmukh@sohamdesh_

we show for the first time ever that sub-billion audio models can reason. we introduce mellow, a small audio-language model (167M) that gets SoTA on different audio reasoning tasks. by using our method and data, you can train an alm within 24 hrs on academic resources (1/n 🧵)

English
1
1
9
719
Mimansa Jaiswal
Mimansa Jaiswal@MimansaJ·
I was impacted by Meta layoffs today. As a Research Scientist working on LLM posttraining (reward models, DPO/GRPO) & automated evaluation pipelines, I’ve focused on understanding why/wehere models fail & how to make them better. I’m looking for opportunities; please reach out!
Susan Zhang@suchenzang

👀

English
121
231
3.1K
841K
Animesh Garg
Animesh Garg@animesh_garg·
Why is credentialed gatekeeping still a thing? @dwarkesh_sp has gained more technical insight, during his deep preparations for the breadth of speakers he meets, than most people would have working full time! And yet, he almost always admits not being the expert in each of his podcasts. Experience is more valuable than pedigree!
taimur@taimurabdaal

I’m intrigued by this Dwarkesh guy but what are his qualifications? E.g. has he done research at a top school like MIT?

English
31
11
432
98.1K
Soham Deshmukh retweetledi
arXiv Sound
arXiv Sound@ArxivSound·
Shikhar Bharadwaj, Samuele Cornell, Kwanghee Choi, Satoru Fukayama, Hye-jin Shim, Soham Deshmukh, Shinji Watanabe, "OpenBEATs: A Fully Open-Source General-Purpose Audio Encoder," arxiv.org/abs/2507.14129
Indonesia
0
7
40
6.5K
Soham Deshmukh retweetledi
Dimitris Papailiopoulos
Dimitris Papailiopoulos@DimitrisPapail·
Some of the most impactful work you can do in academia isn’t cool new algos or novel architectures. It’s data research. Data research isn’t just dumping tokens into a json. It requires a ton of rigorous experimentation, algorithmic thinking, and actually talking to your models.
Ludwig Schmidt@lschmidt3

Very excited to finally release our paper for OpenThoughts! After DataComp and DCLM, this is the third large open dataset my group has been building in collaboration with the DataComp community. This time, the focus is on post-training, specifically reasoning data.

English
4
14
179
15.2K
Anish
Anish@anishmaybe·
That moment when you forget to log "finish" in your @AllTrails app. Still, I'll take the ATH
Anish tweet media
English
3
0
2
129
Soham Deshmukh
Soham Deshmukh@sohamdesh_·
Being presented currently in Hall 3 - poster number 48 - 10 am to 12:30 pm. Come drop by to discuss robustness evaluation for ASR systems! #ICLR2025
Soham Deshmukh tweet media
Ahmed Shah@AhmedSh1494

Happy to announce that Speech Robust Bench (SRB) is being presented at #ICLR2025. SRB is a comprehensive multi-lingual robustness benchmark for speech recognition. paper: openreview.net/forum?id=D0LuQ… code: github.com/ahmedshah1494/… data: huggingface.co/datasets/mshah… more in 🧵

English
0
0
6
317
Soham Deshmukh
Soham Deshmukh@sohamdesh_·
@makemytripcare Absolutely worst customer service! My ticket was confirmed after boarding closed, forcing me to buy another ticket at the counter. After a week of chasing 5 different people, still no updates.
English
1
0
0
24
Soham Deshmukh
Soham Deshmukh@sohamdesh_·
Just landed in Hyderabad for ICASSP 2025! Excited to reconnect with familiar faces and meet some new ones. Feel free to reach out or swing by the Microsoft booth for a quick chat #ICASSP2025
English
0
1
11
867
Soham Deshmukh
Soham Deshmukh@sohamdesh_·
also, if one scales data (audio), mellow performance improves further, highlighting the effectiveness of our minimal recipe. we release the checkpoint, reasonaqa dataset, and will soon open-source the training code to make research on audio-language models more accessible!
English
1
0
2
218
Soham Deshmukh retweetledi
𝚐𝔪𝟾𝚡𝚡𝟾
𝚐𝔪𝟾𝚡𝚡𝟾@gm8xx8·
Mellow: a small audio language model for reasoning
𝚐𝔪𝟾𝚡𝚡𝟾 tweet media
English
2
22
100
8.2K
Soham Deshmukh retweetledi
arXiv Sound
arXiv Sound@ArxivSound·
``Mellow: a small audio language model for reasoning,'' Soham Deshmukh, Satvik Dixit, Rita Singh, Bhiksha Raj, ift.tt/BFgXl2L
Indonesia
0
5
25
3.4K

Keşfet

@HirofumiInaguma @rdesh26 @MimansaJ @animesh_garg @dwarkesh_sp @anishmaybe @AllTrails @makemytripcare