Unverified50% confidenceFactExact time
Recursive self-improvement involves no post-training or gradient updates to the LLM; the model weights remain frozen, and each iteration requires only forward inference.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedLLM采用无状态设计,模型权重在训练后被固化,推理时不会因某次对话而改变内部参数78% similarUnverifiedIn the LLM context, recursive self-improvement need not mean the model directly modifying its own weights; it can take the form of the model proposing improvements at the architecture, data or training recipe level, which are then validated and executed by an external process.77% similarUnverifiedFew-shot prompting does not modify model weights; it constrains the output distribution through context only at inference time, unlike fine-tuning in traditional machine learning.74% similarUnverifiedThe evaluated VLM itself remains completely frozen, and all optimization occurs at the level of the reasoning program and surrounding code, requiring no gradient updates, additional training data, or modification of model weights.74% similarUnverifiedAs long as the model architecture, hyperparameters, data splits, and random seed settings are consistent with the paper's description, retraining the model itself does not constitute academic misconduct.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1154118API
curl https://kongchang.com/api/v1/knowledge/claims/1154118MCP
get_claim(id=1154118)