arkil_patel Profile Banner
Arkil Patel Profile
Arkil Patel

@arkil_patel

Followers
1K
Following
2K
Media
20
Statuses
226

CS PhD Student at Mila and McGill | Worked at AllenNLP and Microsoft Research

Montrรฉal, Quรฉbec
Joined October 2016
Don't wanna be here? Send us removal request.
@arkil_patel
Arkil Patel
9 months
๐“๐ก๐จ๐ฎ๐ ๐ก๐ญ๐จ๐ฅ๐จ๐ ๐ฒ paper is out! ๐Ÿ”ฅ๐Ÿ‹ We study the reasoning chains of DeepSeek-R1 across a variety of tasks and settings and find several surprising and interesting phenomena! Incredible effort by the entire team! ๐ŸŒ: https://t.co/CDlFHD28xQ
@saraveramarjano
Sara Vera Marjanoviฤ‡
9 months
Models like DeepSeek-R1 ๐Ÿ‹ mark a fundamental shift in how LLMs approach complex problems. In our preprint on R1 Thoughtology, we study R1โ€™s reasoning chains across a variety of tasks; investigating its capabilities, limitations, and behaviour. ๐Ÿ”—: https://t.co/Cyy18kYQ45
1
5
26
@bhatia_mehar
Mehar Bhatia
1 month
๐ŸšจHow do LLMs acquire human values?๐Ÿค” We often point to preference optimization. However, in our new work, we trace how and when model values shift during post-training and uncover surprising dynamics. We ask: How do data, algorithms, and their interaction shape model values?๐Ÿงต
2
49
124
@divyat09
Divyat Mahajan
2 months
[1/9] While pretraining data might be hitting a wall, novel methods for modeling it are just getting started! We introduce future summary prediction (FSP), where the model predicts future sequence embeddings to reduce teacher forcing & shortcut learning. ๐Ÿ“ŒPredict a learned
10
47
221
@satwik1729
Satwik Bhattamishra
2 months
Excited to share our new work on the expressivity of Transformer-based multi-agent systems and understanding the trade-offs in communication, no. of agents, and achievable speedups โœจ Work led by @frisbeemortel; check out his thread for details!
@frisbeemortel
Michael Rizvi-Martel
2 months
Is there such a thing as too many agents in multi-agent systems? It depends! ๐Ÿงต Our work reveals 3 distinct regimes where communication patterns differ dramatically. More on our findings below ๐Ÿ‘‡ (1/7)
0
4
13
@a_kazemnejad
Amirhossein Kazemnejad
2 months
Itโ€™s clear next-gen reasoning LLMs will run for millions of tokens. RL at 1M needs ~100ร— compute than 128K. Our Markovian Thinking keeps compute scaling linear instead. Check out Miladโ€™s thread; some of my perspectives below:
@MAghajohari
Milad Aghajohari
2 months
Introducing linear scaling of reasoning: ๐“๐ก๐ž ๐Œ๐š๐ซ๐ค๐จ๐ฏ๐ข๐š๐ง ๐“๐ก๐ข๐ง๐ค๐ž๐ซ Reformulate RL so thinking scales ๐Ž(๐ง) ๐œ๐จ๐ฆ๐ฉ๐ฎ๐ญ๐ž, not O(n^2), with O(1) ๐ฆ๐ž๐ฆ๐จ๐ซ๐ฒ, architecture-agnostic. Train R1-1.5B into a markovian thinker with 96K thought budget, ~2X accuracy ๐Ÿงต
19
93
896
@MAghajohari
Milad Aghajohari
2 months
Introducing linear scaling of reasoning: ๐“๐ก๐ž ๐Œ๐š๐ซ๐ค๐จ๐ฏ๐ข๐š๐ง ๐“๐ก๐ข๐ง๐ค๐ž๐ซ Reformulate RL so thinking scales ๐Ž(๐ง) ๐œ๐จ๐ฆ๐ฉ๐ฎ๐ญ๐ž, not O(n^2), with O(1) ๐ฆ๐ž๐ฆ๐จ๐ซ๐ฒ, architecture-agnostic. Train R1-1.5B into a markovian thinker with 96K thought budget, ~2X accuracy ๐Ÿงต
14
202
919
@arkil_patel
Arkil Patel
2 months
Iโ€™m at CoLM this week! Come check out our work on evaluating RMs for agent trajectories! These days, Iโ€™m thinking about forecasting generalization, scaling laws, and safety/adversarial attacks. Ping me if you wanna chat about research!
@xhluca
Xing Han Lu
2 months
i will be presenting AgentRewardBench at #COLM2025 next week! session: #3 date: wednesday 11am to 1pm poster: #545 come learn more about the paper, my recent works or just chat about anything (montreal, mila, etc.) here's a teaser of my poster :)
0
5
7
@satwik1729
Satwik Bhattamishra
2 months
Check out this new work on techniques for constructing Transformers for algorithmic tasks! Excited to have been part of this project!
@pentagonalize
Andy J Yang
2 months
We present The Transformer Cookbook: a collection of recipes for programming algorithms directly into transformers! Hungry for an induction head? Craving a Dyck language recognizer? We show you step-by-step how to cook up transformers for these algorithms and many more!
0
3
8
@mariusmosbach
Marius Mosbach
3 months
Here's a list of recommendations for what to do in Montreal during @COLM_conf and beyond. ๐Ÿ‘‰: https://t.co/9ixyy2Y7Yl Many thanks to my co-authors @benno_krojer and @frisbeemortel.
Tweet card summary image
github.com
A list of things to do in Montrรฉal. Contribute to mmarius/montreal-things-to-do development by creating an account on GitHub.
@mariusmosbach
Marius Mosbach
3 months
Who will be at @COLM_conf ? I'm preparing a list of recommendations for what to do in beautiful Montreal. Stay tuned. ๐Ÿฅฏ
6
18
64
@Mila_Quebec
Mila - Institut quรฉbรฉcois d'IA
3 months
Exciting news! We're thrilled to announce the appointment of Professor @hugo_larochelle as Mila's new Scientific Director! A deep learning pioneer and former head of Google's AI lab in Montreal, Hugo's leadership will be pivotal in advancing AI for the benefit of all. Read the
12
29
257
@ncmeade
Nicholas Meade
5 months
Come by our #ACL2025 poster tomorrow to discuss the safety risks surrounding increasingly capable instruction-following retrievers (or anything safety related)! 16:00-17:30 on Tuesday in Hall 4/5
@ParishadBehnam
Parishad BehnamGhader
5 months
Come and visit our poster on the Safety of Retrievers @aclmeeting ๐Ÿ—“๏ธTuesday, Findings Posters, 16:00-17:30 ๐ŸšจInstruction-following retrievers will become increasingly good tools for searching for harmful or sensitive information.๐Ÿšจ
0
4
16
@ParishadBehnam
Parishad BehnamGhader
5 months
Come and visit our poster on the Safety of Retrievers @aclmeeting ๐Ÿ—“๏ธTuesday, Findings Posters, 16:00-17:30 ๐ŸšจInstruction-following retrievers will become increasingly good tools for searching for harmful or sensitive information.๐Ÿšจ
@ParishadBehnam
Parishad BehnamGhader
9 months
Instruction-following retrievers can efficiently and accurately search for harmful and sensitive information on the internet! ๐ŸŒ๐Ÿ’ฃ Retrievers need to be aligned too! ๐Ÿšจ๐Ÿšจ๐Ÿšจ Work done with the wonderful @ncmeade and @sivareddyg ๐Ÿ”— https://t.co/yLJPiy1d0j Thread: ๐Ÿงต๐Ÿ‘‡
1
7
19
@mariusmosbach
Marius Mosbach
5 months
@aryopg Nice work! We observed a similar trend on certain math tasks in our work: https://t.co/hNlFcjKauc Section 4.1 has a discussion of our findings. You might want to consider citing it :) cc @saraveramarjano @arkil_patel @sivareddyg
0
5
13
@arkil_patel
Arkil Patel
5 months
If youโ€™re at ICML and if you work on interpretability or causality, go talk to @_shruti_joshi_, she has a fantastic paper!
@_shruti_joshi_
Shruti Joshi
5 months
I will be at the Actionable Interpretability Workshop (@ActInterp, #ICML) presenting *SSAEs* in the East Ballroom A from 1-2pm. Drop by (or send a DM) to chat about (actionable) interpretability, (actionable) identifiability, and everything in between!
0
0
3
@sivareddyg
Siva Reddy
5 months
Come find us at the #ICML2025 poster if you are interested in safety of web agents!
@ncmeade
Nicholas Meade
5 months
I'll be at #ICML2025 this week presenting SafeArena (Wednesday 11AM - 1:30PM in East Exhibition Hall E-701). Come by to chat with me about web agent safety (or anything else safety-related)!
0
5
26
@arkil_patel
Arkil Patel
5 months
SafeArena is being presented at #ICML2025 !! Check out our poster and talk to @ncmeade for all things โ€˜safety โˆช agents โˆช LLMsโ€™!
@ncmeade
Nicholas Meade
5 months
I'll be at #ICML2025 this week presenting SafeArena (Wednesday 11AM - 1:30PM in East Exhibition Hall E-701). Come by to chat with me about web agent safety (or anything else safety-related)!
0
1
11
@arkil_patel
Arkil Patel
6 months
Congrats @vernadankers!! Weโ€™re lucky to have you join our lab!
@tallinzen
Tal Linzen
6 months
Congratulations Verna! This was one of the best theses I've ever read, I highly recommend checking out Verna's work on the tradeoffs between memorization and generalization in language models!
0
0
5
@vernadankers
Verna Dankers
6 months
I miss Edinburgh and its wonderful people already!! Thanks to @tallinzen and @PontiEdoardo for inspiring discussions during the viva! I'm now exchanging Arthur's Seat for Mont Royal to join @sivareddyg's wonderful lab @Mila_Quebec ๐Ÿคฉ
@agostina_cal
Agostina Calabrese ๐Ÿฆ‹
6 months
Huge congratulations to Dr. @vernadankers for passing her viva today! ๐Ÿฅณ๐ŸŽ“ It's been an honour sharing the PhD journey with you. I wasnโ€™t ready for the void your sudden departure left (in the office and in my life!). Your new colleagues are lucky to have you! ๐Ÿฅบ๐Ÿฅฐ @Edin_CDT_NLP
11
11
100
@xhluca
Xing Han Lu
6 months
"Build the web for agents, not agents for the web" This position paper argues that rather than forcing web agents to adapt to UIs designed for humans, we should develop a new interface optimized for web agents, which we call Agentic Web Interface (AWI).
9
59
197
@ziling_cheng
Ziling Cheng @ EMNLP
6 months
Do LLMs hallucinate randomly? Not quite. Our #ACL2025 (Main) paper shows that hallucinations under irrelevant contexts follow a systematic failure mode โ€” revealing how LLMs generalize using abstract classes + context cues, albeit unreliably. ๐Ÿ“Ž Paper: https://t.co/YEK4TaI7pq 1/n
6
26
44
@kabirahuja004
Kabir
8 months
๐Ÿ“ข New Paper! Tired ๐Ÿ˜ด of reasoning benchmarks full of math & code? In our work we consider the problem of reasoning for plot holes in stories -- inconsistencies in a storyline that break the internal logic or rules of a storyโ€™s world ๐ŸŒŽ W/ @melaniesclar, and @tsvetshop 1/n
3
54
262