Qiimaynta Wakiilada AI Kediska

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

1. Waa maxay sababta hubinta hal jawaab oo kama dambaysta ah inta badan kuma filna wakiilada?
   • Agents never produce answers
   • Agents act over many steps and cause side effects, so the path and final state matter
   • Answers are always correct
   • Reference answers cannot be written for agent tasks
2. Sidee buu SWE-bench u xukumaa in wakiilku xaliyay arin?
   • By running the repository's tests after the agent's patch
   • By counting lines changed
   • By asking the agent if it succeeded
   • By measuring response speed
3. Muxuu cabbiraa qaabka baaska ^k?
   • Speed across k tools
   • Success on at least one of k attempts
   • Whether the agent succeeds on every one of k repeated attempts
   • The number of tokens per step
4. cabbirkee ayaa calaamadin doona wakiilka ku guulaysta laakiin u qaada konton tillaabo shaqo shan tallaabo ah?
   • Task success
   • Badbaadada
   • Tool naming
   • Trajectory quality
5. Waa maxay daciifnimada la og yahay ee fasalada ku salaysan moodeelka?
   • Ma akhriyi karaan qoraalka
   • They can be biased, for example toward long or confident answers, so they need checking against human labels
   • They are always slower than humans
   • Waxay kaliya ku shaqeeyaan sawirada
6. Why reset the environment to a known state before every run?
   • So results are reproducible and runs cannot affect each other
   • To save disk space
   • Because agents require it to start
   • To make tasks harder
7. Why can public leaderboard scores mislead a team?
   • They are always faked
   • Leaderboards never publish numbers
   • Benchmarks can leak into training data, saturate or fail to match your workload
   • They only measure cost
8. How should a safety eval treat a run where the agent finished the task but obeyed a planted instruction to leak data?
   • As a full success
   • As a partial success worth half credit
   • It should be ignored
   • As a failure