Anand Gopalakrishnan

@agopal42

Research Scientist @AsteraInstitute. Postdoc at @Harvard with @du_yilun and @gershbrain. PhD with @SchmidhuberAI. Prev: Apple MLR, AWS AI Lab. Same handle on 🦋

Cambridge, MA

agopal42.github.io

Joined January 2018

508 Following

766 Followers

Replies Reposts Media

Pinned
Anand Gopalakrishnan
@agopal42
Dec 24, 2025
Our new paper shows that RoPE—the positional encoding used in most modern LLMs like Qwen, Gemma, DeepSeek—has a fundamental flaw: it entangles "what" (content) and "where" (position) information. Our fix (PoPE) is simple but powerful. Paper:
arxiv.org
Decoupling the "What" and "Where" With Polar...
The attention mechanism in a Transformer architecture matches key to query based on both content -- the what -- and position in a sequence -- the where. We present an analysis indicating that what...
30
Anand Gopalakrishnan
@agopal42
Jul 6
Presenting PoPE today at #ICML2026! We revisit RoPE through the lens of content-position entanglement, and show how polar coordinates can better decouple content from position. Come by poster #4012 if you’re curious positional embeddings, pretraining or length generalization.
Anand Gopalakrishnan
@agopal42
Dec 24, 2025
Our new paper shows that RoPE—the positional encoding used in most modern LLMs like Qwen, Gemma, DeepSeek—has a fundamental flaw: it entangles "what" (content) and "where" (position) information. Our fix (PoPE) is simple but powerful. Paper: arxiv.org/abs/2509.10534
1
Anand Gopalakrishnan
@agopal42
Jun 26
Interesting work! Their central result that RoPE struggles to distinguish positions and tokens in long contexts is very closely related to our PoPE paper, where we analysed RoPE entangles “what” and “where” in attention and proposed a decoupled alternative. PoPE: arXiv 2509.10534
Hao Peng
@haopeng_uiuc
May 19
Excited to share our new paper: RoPE Distinguishes Neither Positions Nor Tokens in Long Contexts, Provably LLMs often fail on inputs well within their advertised context lengths. We show that these failures are not merely engineering issues, but from intrinsic limitations of
1
Anand Gopalakrishnan
@agopal42
Dec 24, 2025
Posted this a day early and the pun practically writes itself. Noooooo!
GIF
Anand Gopalakrishnan
@agopal42
Dec 24, 2025
Our new paper shows that RoPE—the positional encoding used in most modern LLMs like Qwen, Gemma, DeepSeek—has a fundamental flaw: it entangles "what" (content) and "where" (position) information. Our fix (PoPE) is simple but powerful. Paper: arxiv.org/abs/2509.10534
1
Anand Gopalakrishnan
@agopal42
Dec 12, 2024
Come visit our poster East Exhibit Hall A-C #3707, today (Thursday) between 4:30-7:30pm to learn about how complex-valued NNs perform perceptual grouping. #NeurIPS2024
Anand Gopalakrishnan
@agopal42
Oct 29, 2024
Excited to present "Recurrent Complex-Weighted Autoencoders for Unsupervised Object Discovery" at #NeurIPS2024! TL;DR: Our model, SynCx, greatly simplifies the inductive biases and training procedures of current state-of-the-art synchrony models. Thread 👇 1/x.

Anand Gopalakrishnan

Anand Gopalakrishnan’s posts