LINKED LIST
[txt mode]
▸
A small number of samples can poison ...
home
explore
|
log in
A small number of samples can poison LLMs of any size
anthropic.com · first added by
@hnl
· 2026-02-27 · 3 upvotes
log in
to save, upvote or flag this.
─── In 2 lists ─────────────────────────────────────────
*
Issue #766
by
@hnl
[data]
*
Issues
by
@tech_daily
related read on:
*
Why your boss isn't worried about AI
(from the discussion)
─── Discussions ────────────────────────────────────────
*
A small number of samples can poison LLMs of any size
1202 pts · 439 comments ·
node
─── From the discussion ────────────────────────────────
*
Clankers Die on Christmas
remyhax.xyz ·
node
*
Data Poisoning in LLMs: Jailbreak-Tuning and Scaling Laws
arxiv.org ·
node
*
AI Can Write Code Like Humans—Bugs and All
wired.com ·
node
*
Does your LLM truly unlearn? An embarrassingly simple approach to recover unlearned knowledge
arxiv.org ·
node
*
https://www.washingtonpost.com/technology/2025/08/15/google-ai-overviews-scam/
washingtonpost.com ·
node
see all 6
*
Metadata Conditioning Accelerates Language Model Pre-training
arxiv.org ·
node
─── Related hn threads ─────────────────────────────────
*
Clankers Die on Christmas
269 pts · 249 comments ·
node
─── Also saved alongside this ──────────────────────────
*
AI makes you boring
6 upvotes · 2026-02-23
*
Isometric NYC
4 upvotes · 2026-01-26
*
ooh.directory: a place to find good blogs that interest you
7 upvotes · 2026-01-31
*
pi.dev
6 upvotes · 2026-02-06