RSS

Sohaib Imran

Karma: 168

Research associate at the University of Lancaster, studying AI safety.

My current views on AI:


[Question] Does the ChatGPT (web)app some­times show ac­tual o1 CoTs now?

Sohaib Imran29 Jan 2025 17:27 UTC
6 points
6 comments1 min readLW link

Cross-con­text ab­duc­tion: LLMs make in­fer­ences about pro­ce­du­ral train­ing data lev­er­ag­ing declar­a­tive facts in ear­lier train­ing data

Sohaib Imran16 Nov 2024 23:22 UTC
36 points
11 comments14 min readLW link

Arte­facts gen­er­ated by mode col­lapse in GPT-4 Turbo serve as ad­ver­sar­ial at­tacks.

Sohaib Imran10 Nov 2023 15:23 UTC
11 points
0 comments2 min readLW link

Introduction

30 Jun 2023 20:45 UTC
8 points
0 comments2 min readLW link

In­her­ently In­ter­pretable Architectures

30 Jun 2023 20:43 UTC
4 points
0 comments7 min readLW link

Pos­i­tive Attractors

30 Jun 2023 20:43 UTC
6 points
0 comments13 min readLW link

AISC 2023, Progress Re­port for March: Team In­ter­pretable Architectures

2 Apr 2023 16:19 UTC
14 points
0 comments14 min readLW link