Mechanistic Interpretability Quiz

Mhinduro uye tsananguro zvinoonekwa paunenge uchitora bvunzo.

1. Chii chiri chinangwa chepakati chekududzira mechanistic?
   • Kudzoreredza-injiniya network yemukati makomputa kuita munhu-anonzwisisika algorithms
   • Kuita mamodheru kudzidzisa nekukurumidza
   • Kumanikidza mamodheru kune nharembozha
   • Kuisa zvimiro zvekupinza nekukosha chete
2. Chii chinonzi 'superposition' chinoreva mune neural network?
   • Kubatanidza akawanda mamodheru pamwechete
   • Rinomiririra zvinhu zvakawanda kupfuura zvimiro nekushandisa mafambiro anopindirana
   • Kumhanyisa modhi pane akawanda maGPU
   • Avhareji yekufungidzira kubva kune dzakasiyana kumhanya
3. Sparse autoencoder anoshandiswa mukududzira mechanistic kunyanya ku...
   • Gadzira data yekudzidziswa
   • Kurumidza kufungidzira
   • Disentangle polysemantic activation kuita yakachena, inodudzirwa zvakanyanya
   • Encrypt modhi uremu
4. Chii chinonzi 'induction heads'?
   • Hardware accelerators kune transformers
   • Mitsara inobuda inoburitsa zvinyorwa
   • Mhando yebasa rekurasikirwa
   • Misoro yekutarisa inokopa uye inoenderera ichidzokororwa tokeni mapatani, ichitsigira mu-context kudzidza
5. Vatsvakurudzi vanosimbisa sei kuti dunhu rakawanikwa rinonyatsoita computation yekufungidzira?
   • Kuburikidza necausal kupindira senge kubvisa kana kupeta ma activation chaiwo
   • Nokuongorora chokwadi chekudzidzisa kwemuenzaniso
   • Nekuverenga nhamba yeparameters
   • Nekuyera inference latency
6. What is a cautious safety-related reason to study mechanistic interpretability?
   • It proves every hidden behavior can already be found
   • It may provide evidence about internal computations, while current methods are not a complete audit or direct steering guarantee
   • It guarantees a model will follow any requested correction
   • It replaces evaluations of model behavior in deployment
7. If a sparse autoencoder feature activates on examples described as “Golden Gate Bridge,” what does that observation establish?
   • The entire model’s computation is now explained
   • The feature is guaranteed to be a single neuron
   • A candidate feature correlates with those examples and needs further testing to establish its role
   • The feature causes every bridge-related output
8. Why test a candidate feature across varied prompts and examples?
   • Proof that the model has no other relevant circuits
   • A guarantee that the explanation generalizes to every prompt
   • A replacement for inspecting model behavior
   • To check whether the observed activation pattern generalizes beyond selected examples