Unverified50% confidenceFactExact time
TS-ICL focuses on in-context learning for time series, that is, enabling a model to complete predictions for new tasks based solely on the input context without updating parameters.
1
Sources
50%
Confidence
Long-term
Relevance
10/18/2026
First Seen
Sources
TempoPFN: Synthetic Pre-training of Linear RNNs Enables Zero-Shot Time Series Forecasting
redditr/MachineLearning10/18/2026
Related Entities
Related Claims
UnverifiedAction chunk value learning lets the policy predict a sequence of actions within a time window (action chunk) all at once, and estimate the value of this chunk as a whole, rather than outputting single-step actions frame by frame.71% similarUnverifiedTest-Time Adaptation (TTA) allows a model to perform online updates of batch normalization layer parameters using a small amount of target domain data during inference, serving as a lightweight approach to addressing domain shift.69% similarVerifiedIn-context learning (ICL) refers to the ability of large language models to perform new tasks without updating weights, simply by providing examples or instructions in the input prompt, a capability first systematically described in the GPT-3 paper69% similarUnverifiedLarge model inference refers to the process by which a model, after completing training, receives user input and generates output in real time, and is extremely sensitive to latency.68% similarUnverifiedInference-time control refers to the technique of guiding a pretrained generative model toward a specific target distribution without modifying its parameters.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1062818API
curl https://kongchang.com/api/v1/knowledge/claims/1062818MCP
get_claim(id=1062818)