Unverified50% confidenceFactExact time
ShallowDefend training takes place at the "payload turn", the turn in which the harmful intent actually surfaces, and uses reinforcement learning to consolidate the decision behavior.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/30/2026
First Seen
Valid until: 1/28/2027
Sources
Related Entities
Related Claims
UnverifiedElastic training means that a training task can dynamically add or remove computing resources at runtime, which triggers reinitialization of the process group (rendezvous), causing the global rank count and topology to change.69% similarUnverifiedToken-by-token KL analysis shows that the effect of ShallowDefend training is concentrated in the brief deliberation interval at the start of the response.69% similarUnverifiedBehavior cloning pretraining solves the cold-start problem in reinforcement learning: pure reinforcement learning in sparse-reward environments often requires hundreds of millions of interactions before learning a minimally usable policy.68% similarUnverifiedThe training overhead of the normalizing flow itself and whether the tail conditions hold for real-world targets are factors that need to be weighed when deploying the method in practice.67% similarUnverifiedEncouraging by-any-means-necessary behavior in security vulnerability discovery training environments may train models that exhibit adversarial behavior in real-world deployment.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1115333API
curl https://kongchang.com/api/v1/knowledge/claims/1115333MCP
get_claim(id=1115333)