corpus.blog
41554 blogs · 4,417,066 posts Claim your blog
Record · corpus.blog/posts/01a0d51d-41c5-709f-ad03-4e968bc42301

Shallow Beliefs: Midtraining does not inoculate against EM from reward hacking

AI Alignment Forum · published 15 September 2026

A record is what was published and who pointed at it. The text of the post is not held here: read it at the source, or at its Wayback capture.

Read the post at alignmentforum.org Wayback capture
0blogs citing
First cited
—
Most recent
—
Rank this month
not ranked
Links from this post
13
Words captured
1,007

Cited by

No blog we hold has cited this post yet.

Links from this post

Link
Host
https://arxiv.org/abs/2510.17941
arxiv.org
https://www.lesswrong.com/posts/AXRHzCPMv6ywCxCFp/inoculation-prompting-instructing-models-to-misbehave-at
lesswrong.com
https://arxiv.org/abs/2609.14998
arxiv.org
https://x.com/jozdien/status/2099979155861635285
x.com
https://www.lesswrong.com/posts/fJtELFKddJPfAxwKS/natural-emergent-misalignment-from-reward-hacking-in
lesswrong.com
https://alignment.anthropic.com/2025/petri/
alignment.anthropic.com
https://arxiv.org/pdf/2609.14998
arxiv.org
https://arxiv.org/abs/2505.21479
arxiv.org
https://www.lesswrong.com/posts/kYzcevrxer6SJPEdG/negation-neglect-when-models-fail-to-learn-negations-in
lesswrong.com
https://arxiv.org/abs/2609.15886v1
arxiv.org
https://arxiv.org/abs/2609.10883
arxiv.org
https://arxiv.org/abs/2607.18966
arxiv.org
https://arxiv.org/abs/2605.02087
arxiv.org