Unverified50% confidenceFactExact time
Goal injection works by providing the agent or simulator with additional partial information about the target item during training (such as category and attribute keywords), artificially increasing the reachability of the task, and it is used only during training and completely removed at inference.
1
Sources
50%
Confidence
Long-term
Relevance
11/13/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedProgressive disclosure means that the Agent gradually loads and discloses relevant skill information at runtime according to task needs, rather than exposing all capabilities to the model at once.76% similarUnverifiedIn the machine learning context, HITL primarily referred to human-annotated data participating in a model training feedback loop, and has been redefined in Agent workflows as runtime intervention.73% similarUnverifiedLLM agent skill distillation refers to extracting reusable experience from the historical trajectories of an agent performing tasks and solidifying it into a skill representation that can be directly invoked for subsequent tasks without updating model parameters.72% similarVerifiedCommon methods for constructing Agent Tuning training data include trajectory distillation, self-play, and manual annotation correction72% similarVerifiedAgent Skill adopts a lazy-loading mechanism whereby the Agent first reads only the metadata (name and description) of each Skill and loads the full instructions only after matching the user's need, so having multiple Skills coexisting does not increase Token consumption71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1165061API
curl https://kongchang.com/api/v1/knowledge/claims/1165061MCP
get_claim(id=1165061)