Unverified50% confidenceFactExact time
The core operation of RC-DS is: first perform a singular value decomposition on the readout transformation, remove the singular value gains, keep only the orthogonal or rotational part that determines the directional structure, and then train the SAE in the normalized decision-space coordinates.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSDAE uses reflection tokens to locate which steps have a critical influence on the execution of a given subtask, directing gradients to concentrate at those positions.65% similarUnverifiedDSpec引入微小记忆头（memory head）防止草稿模型中途崩坏，并引入动态调度器智能选择需验证的Token，实现无损加速64% similarUnverifiedThe core innovation of DSAS lies in decoupling 'when to steer' from 'how to steer': DSAS dynamically determines when to apply intervention, while the direction and form of the steering transformation still follow existing methods.64% similarUnverifiedDSAS can adaptively and dynamically adjust steering intensity across different layers and different inputs, with intervention intensity approaching zero when the input is normal and amplifying intervention when a potentially harmful tendency is detected.62% similarUnverifiedImplementing a Transformer from scratch requires understanding numerical stability treatments such as why attention scores are divided by √d_k, and is a watershed task that distinguishes 'knowing how to use frameworks' from 'understanding the underlying principles'.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1023678API
curl https://kongchang.com/api/v1/knowledge/claims/1023678MCP
get_claim(id=1023678)