Self-Consistency Decoding Quiz

Mhinduro uye tsananguro zvinoonekwa paunenge uchitora bvunzo.

1. How does self-consistency decoding pick a final answer from a language model?
   • Enzanisa nzira dzakawanda dzekufunga uye tora vhoti yakawanda pamhinduro
   • Nguva dzose sarudza chiratidzo chimwe chete chinogona kuitika padanho rega rega
   • Dzidzisa modhi pane zvayaburitsa kakawanda
   • Bvisa zvese zvisina tsarukano kubva kuchizvarwa
2. Under what condition can self-consistency improve answer accuracy?
   • Whenever the number of samples increases
   • When correct sampled paths converge and errors are not too strongly correlated
   • When every reasoning chain has identical wording
   • Only when the model weights are retrained
3. Ndeipi nzira yekudhirodha inotsiva yega-inotsiva?
   • Kudzokera kumashure
   • Kukamura Mazwi
   • Kukara decoding yeketani imwe chete
   • Weight quantization
4. Ndeipi chikamu checheni yega yega yemuenzaniso chinonyatso shandiswa muvhoti yekupedzisira?
   • Rugwaro rwakazara rwekufunga
   • Nhamba yezviratidzo zvakagadzirwa
   • The temperature setting
   • Mhinduro yekupedzisira chete
5. What is the main cost of increasing the number of self-consistency samples?
   • More inference compute; accuracy gains depend on the task and can have diminishing returns
   • It retrains the model weights after every response
   • It guarantees that the majority is correct
   • It eliminates the need to sample any reasoning paths
6. Kuzvimiririra muenzaniso wekutanga weiyo pfungwa yakakura?
   • Model compression
   • Inference-nguva kuyera (kushandisa imwe compute kuti ufunge zviri nani)
   • Kuwedzera data panguva yekudzidziswa
   • Hardware acceleration
7. Why may answer normalization be useful before a self-consistency vote?
   • It makes all reasoning paths independent
   • It verifies the answer against external evidence
   • It can group semantically equivalent answers that use different formatting
   • It changes the model’s sampled probabilities
8. What is a key limitation of a majority answer from self-consistency?
   • A majority vote cannot be computed for text
   • Sampling always changes the model weights
   • Every reasoning path is guaranteed independent
   • Correlated reasoning errors can make an incorrect answer win the vote