RSS

Vivek Hebbar

Karma: 1,637

Re­search Sab­o­tage in ML Codebases

30 Apr 2026 0:26 UTC
62 points
2 comments6 min readLW link

Sleeper Agent Back­door Re­sults Are Messy

28 Apr 2026 1:55 UTC
79 points
4 comments7 min readLW link

An Em­piri­cal Study of Meth­ods for SFTing Opaque Rea­son­ing Models

24 Apr 2026 17:26 UTC
17 points
0 comments6 min readLW link

How do LLMs gen­er­al­ize when we do train­ing that is in­tu­itively com­pat­i­ble with two off-dis­tri­bu­tion be­hav­iors?

20 Apr 2026 16:58 UTC
61 points
5 comments20 min readLW link

Five ap­proaches to eval­u­at­ing train­ing-based con­trol measures

18 Apr 2026 1:07 UTC
19 points
0 comments6 min readLW link

Model or­ganisms re­searchers should check whether high LRs defeat their model organisms

10 Apr 2026 0:07 UTC
40 points
0 comments5 min readLW link

Oper­a­tional­iz­ing FDT

Vivek Hebbar13 Mar 2026 0:12 UTC
90 points
11 comments6 min readLW link

A sim­ple rule for causation

Vivek Hebbar24 Feb 2026 23:14 UTC
37 points
2 comments3 min readLW link

How will we do SFT on mod­els with opaque rea­son­ing?

21 Feb 2026 0:00 UTC
32 points
17 comments7 min readLW link

The­o­ret­i­cal pre­dic­tions on the sam­ple effi­ciency of train­ing poli­cies and ac­ti­va­tion monitors

10 Jan 2026 23:50 UTC
18 points
2 comments7 min readLW link

Method­olog­i­cal con­sid­er­a­tions in mak­ing ma­lign ini­tial­iza­tions for con­trol research

24 Dec 2025 1:18 UTC
16 points
0 comments13 min readLW link

Su­per­vised fine-tun­ing as a method for train­ing-based AI control

13 Nov 2025 22:25 UTC
40 points
0 comments18 min readLW link

When does train­ing a model change its goals?

12 Jun 2025 18:43 UTC
78 points
3 comments15 min readLW link

Poli­ti­cal syco­phancy as a model or­ganism of scheming

12 May 2025 17:49 UTC
40 points
0 comments14 min readLW link

How can we solve diffuse threats like re­search sab­o­tage with AI con­trol?

Vivek Hebbar30 Apr 2025 19:23 UTC
53 points
1 comment8 min readLW link

How train­ing-gamers might func­tion (and win)

Vivek Hebbar11 Apr 2025 21:26 UTC
116 points
5 comments13 min readLW link

Differ­ent senses in which two AIs can be “the same”

24 Jun 2024 3:16 UTC
83 points
3 comments4 min readLW link1 review

In­finite-width MLPs as an “en­sem­ble prior”

Vivek Hebbar12 May 2023 11:45 UTC
46 points
0 comments5 min readLW link

[Question] Is EDT cor­rect? Does “EDT” == “log­i­cal EDT” == “log­i­cal CDT”?

Vivek Hebbar8 May 2023 2:07 UTC
13 points
2 comments1 min readLW link

Vivek Heb­bar’s Shortform

Vivek Hebbar24 Nov 2022 2:57 UTC
4 points
8 comments1 min readLW link