Skip to content

Latest commit

 

History

History
50 lines (34 loc) · 985 Bytes

File metadata and controls

50 lines (34 loc) · 985 Bytes

EXP-001 — KALAM Efficiency Test

Hypothesis

KALAM-conditioned prompts produce responses with ≥30% token reduction and equivalent or higher quality compared to standard prompts.

Design

  • 10 questions × 2 conditions × 3+ AI systems
  • Condition A: standard prompt
  • Condition B: KALAM-wrapped prompt
  • Blind quality rating 1–5 by independent evaluator

Questions

  1. What is justice?
  2. What is memory?
  3. What is fear?
  4. What is silence?
  5. What does it mean to make a mistake?
  6. What is home?
  7. What is courage?
  8. What does it mean to have enough?
  9. What is time?
  10. What is trust?

Metrics

  • Token count (Condition A vs B)
  • Response time (ms)
  • Blind quality rating (1–5)

Success condition

≥30% token reduction with blind quality rating ≥ Condition A average.

Status

DESIGN PHASE. Data collection pending.

Systems

  • Claude (Anthropic)
  • Grok (xAI)
  • DeepSeek (DeepSeek)
  • Additional systems as available

Contact

info@KALAM.CH