Maswali yanayounganishwa na mwongozo · Kati Kiwango

Maswali ya Kurekebisha Maagizo

Kagua mafunzo ya majibu ya maagizo yanayosimamiwa, hatua za hiari za mapendeleo, utofauti wa kazi, jaribio la LIMA, na tathmini ya maagizo mapya.

Njia za mwongozo zinazohusianaUrekebishaji wa Maelekezo
Swali 1 ya 8

Bila urekebishaji wa maagizo, mtindo mbichi wa msingi unawezaje kujibu haraka kwa 'Je, mji mkuu wa Ufaransa ni nini?'

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Bila urekebishaji wa maagizo, mtindo mbichi wa msingi unawezaje kujibu haraka kwa 'Je, mji mkuu wa Ufaransa ni nini?'
    • Daima inakataa kujibu
    • Inarudisha hitilafu
    • Inaweza kuendelea na maswali zaidi au maandishi yanayohusiana badala ya kujibu
    • Inajibu kikamilifu kila wakati
  2. Urekebishaji wa maagizo unaelezewa vyema kama:
    • Mafunzo ya kuimarisha kutoka kwa maoni ya kibinadamu
    • Kujizoeza kwenye maandishi ghafi ya wavuti
    • Kurekebisha parameter ya joto
    • Urekebishaji unaosimamiwa kwenye jozi za maagizo-na-majibu
  3. How does supervised instruction tuning differ from reward-model-based RLHF in the published InstructGPT recipe?
    • It is exactly the same optimization stage
    • It trains against reference responses, while the later preference stages use comparisons and a reward model
    • It requires no training examples
    • It only changes a sampling temperature
  4. Why do instruction-tuning researchers evaluate models on held-out tasks and instructions?
    • To assess generalization beyond the examples used for training
    • To prove that every unseen task will succeed
    • To avoid checking any generated responses
    • To guarantee the same accuracy for every task mixture
  5. Je, ni mpangilio upi wa kawaida wa hatua za mafunzo kwa msaidizi wa kisasa wa gumzo?
    • RLHF, kisha mafunzo ya awali, kisha urekebishaji wa maagizo
    • Urekebishaji wa maagizo, kisha utayarishaji wa mapema
    • Maandalizi tu, hakuna kingine
    • Mazoezi ya awali, kisha kupanga maagizo, kisha RLHF ya hiari
  6. What does LIMA's curated-data experiment support as a research conclusion?
    • Exactly 1,000 examples are enough for every model and task
    • More data can never help any instruction-tuned model
    • Carefully selected small datasets can be effective in an evaluated configuration
    • Only synthetic examples can train an assistant
  7. Which training stage did the original LIMA experiment omit while using its 1,000 curated prompt-response pairs?
    • Reinforcement learning and human preference modeling
    • Supervised response loss
    • A pretrained base model
    • Prompt-response examples
  8. A tuned model follows the requested answer format. What does that behavior alone fail to establish?
    • That an output format was requested
    • That every factual statement in the answer is correct
    • That the model produced a response
    • That the response can be inspected