Amit Kumar Jaiswal

@natashamalpani @SarvamAI That's a useful step forward. Could you share the open source model weights?

today is an important day. @SarvamAI just released two open-source foundation models trained from scratch: 30B and 105B parameters. the weights are public. which means the world can verify. that alone matters. training at 100B scale requires distributed systems competence, data pipelines, evaluation frameworks and sustained engineering depth. india now has a team that has done it end-to-end. but the models are only part of the story. what sarvam has actually built this year is a stack. speech recognition tuned for indian languages and code-switching, with reported performance that is competitive on indian benchmarks. text-to-speech optimized for telephony environments. OCR built for indian scripts and messy document formats. on-device models designed to run offline in low-connectivity contexts. and now frontier-scale models sitting at the top of that pyramid. this is a serious attempt to build AI infrastructure for india’s linguistic and distribution reality. some caveats are real given the early stage: independent standardized benchmarking is still emerging. deployment scale is not public. but capability is the point. deepseek showed that frontier research does not require an american zip code. today sarvam showed that frontier training capability exists in india.

@michahu8 Well similar case for a paper in my slot as well, just inspecting the results in case if it interprets and informs their methodologies.

i'm reviewing a paper for ICML that clearly came out of a slop cannon and yet i cannot use LMs while reviewing?? what are we even doing here people

You might look into this as if this is how it operates, one wouldn't want to use your service from next time. Complaint number 53888.

Have been a plus member of @pharmeasyapp. I did an order earlier today where one of the medicine in the list was wrongly assigned as the one selected, and your support agent saying to wait until i receive the order to return, leading to cancel it entirely.

⏰ DHOW-MiLLA #TheWebConf2026 Deadline extended! Call for Papers 👇 Now due: 7 January 2026 Venue: Dubai, UAE 🔗 Submit: openreview.net/group?id=ACM.o… More details: dhow-workshop.github.io #misinformation #factcheck #llm #AI #ML #web #nlp

Deadline: December 18, 2025 (AoE) Please consider submitting your best work. We already have an amazing lineup of invited speakers and the program will get even more exciting with contributed presentations. Join us in Dubai @ ACM #WWW2026

Focus areas are #misinformation detection, computational fact-checking, #GenAI in mitigation of harmful contents, #deepfake, data poisoning and inadvertent harmful data transfer in #LLM-centric/Agentic systems, and many other exciting topics.

🔬 Excited to announce #TheWebConf2026 Joint Workshop on DHOW-MiLLA: Diffusion of Harmful Content on Online Web and Countering Misinformation in the Age of LLMs and Agents. Website: dhow-workshop.github.io Submissions via Openreview: openreview.net/group?id=ACM.o… More details👇

@SharonYixuanLi @iclr_conf They should be, on the contrary @aistats_conf is doing a fantastic job on notifying to everyone who's involved.

I call @iclr_conf to send an official email to ALL authors & reviewers regarding the leakage incident: - We need transparency in communication. A conference cannot rely on social media; a direct email ensures everyone gets the same facts. - We need clear guidance on the remediation & protocols going forward (how fairness will be maintained, options for reviewer recusal) This is essential to prevent a breakdown in community confidence.

This was my first submission at #ICLR and it turns out to be 2/3 of reviews fully AI-generated. Thanks to @pangramlabs @bradley_emi @max_spero_ @gneubig for this super useful tool, yet zemblanity is unavoidable.

@Branchini_Nic Like AISTATS has J2C, and similarily AAAI/IJCAI too

But hey, the *prestige* of ICLR w.r.t. TMLR is so strong that we need a J2C eh ...

🎯 Focusing on Twitter/X social media activity, participants are tasked with predicting the level of #misinformation associated with tweets made by users. This shared task underscores the increasing importance of #machinelearning #AI.

📣 Shared Task on Misinformation Detection in Social Media Launch Alert! This task is hosted on #Codabench within PROMID track @fire_forum 🎯 Build a model for the shared task and submit your predictions. ⏰ Deadline: October 15, 2025 Enroll & learn at codabench.org/competitions/1…

@EzgiKorkmazAI @NeurIPSConf Congratulations!

✨My single author paper just got accepted as a spotlight presentation at @NeurIPSConf. Truly excited! More news to come! #NeurIPS2025

@kchonyc Well, that's kind of becoming a major factor for small groups/solo authors to decide for submitting into JMLR due to long wait, which could lead to substantial new addition until it gets public.

when you give up on this nebulous idea and illusion of prestige, you will finally find peace and freedom. submit to TMLR and JMLR.

@JosephJacks_ Intent is right provided one aware of the underlying abstraction layer

Bitcoin is a single Bittensor subnet.

@jm_alexia Can you share a bit more on the underlying architectures? Or is it via compression?

I'm working on super tiny neural networks that perform better than the big ones, it's kinda insane.

A quick reminder: Workshop on Graph-Augmented LLMs (GaLM) at @ICDM2025 is still accepting submissions! #LLM #GraphML #GraphAnalytics 📝 Papers — Extended deadline: September 5 More Info to submit your work here: iitbhu.ac.in/cf/jcsic/activ…

We've extended the submission deadline of our #IEEE ICDM Workshop on Graph-Augmented LLMs: Bridging Language and Structured Knowledge to September 5th (AoE)! For more info on the workshop and how to submit your work: iitbhu.ac.in/cf/jcsic/activ… #LLM #AI #GraphML #RAG #Analytics

📌 Topics: Graph-enhanced LLM architectures, Multimodal graph #LLM, Graph #RAG, real-world applications and more. #AI

⏰ Deadline: 29th Aug'25 ✨️ Workshop on Graph-Augmented LLMs (GaLM): Bridging Language and Structured Knowledge at #IEEE #ICDM Call for Papers: lnkd.in/esbPm2r8 💡 IEEE ICDM Workshop Proceedings -> Submit here and still publish at top-tier venues!