Kuis Penggabungan Model

Jawaban dan penjelasan muncul saat Anda mengikuti kuis.

1. Apa yang terutama digabungkan dengan penggabungan model?
   • Bobot (parameter) beberapa model
   • Kumpulan data pelatihan beberapa model
   • Prediksi keluaran beberapa model pada waktu inferensi
   • Akselerator perangkat keras yang digunakan untuk pelatihan
2. 'Vektor tugas' paling tepat digambarkan sebagai:
   • Daftar tugas yang dapat dilakukan model
   • Perbedaan antara bobot model yang disetel dengan baik dan bobot model dasarnya
   • Vektor token masukan untuk suatu tugas
   • Gradien dihitung selama satu langkah pelatihan
3. In the described weight-space merging methods, what can be avoided?
   • Evaluation of the merged checkpoint
   • Compatibility requirements between model weights
   • Additional gradient training and access to the original training data
   • Testing the result on intended tasks
4. How do TIES-Merging and DARE differ in addressing parameter interference?
   • Both use the same sign-voting method and differ only in scale
   • DARE uses prompt engineering while TIES retrains on joint data
   • Both average model outputs instead of changing weights
   • They use different techniques: TIES resolves sign conflicts, while DARE drops and rescales task-vector entries
5. Prasyarat utama agar dua model dapat digabungkan dengan sukses adalah keduanya:
   • Berbagi arsitektur yang sama dan biasanya diturunkan dari model dasar yang sama
   • Dilatih pada data yang identik
   • Miliki jumlah langkah pelatihan yang persis sama
   • Gunakan ukuran kosakata tokenizer yang sama saja
6. In task arithmetic, what do coefficients control when combining task vectors?
   • Which tokenizer each checkpoint must use
   • How strongly each parameter-change vector contributes to the merged checkpoint
   • How many training examples are generated
   • Whether an evaluation set is required
7. What are the three main stages named by TIES-Merging?
   • Tokenize text, infer emotions, and export audio
   • Quantize weights, shard GPUs, and route network packets
   • Trim small changes, elect a sign direction, then merge aligned parameter updates
   • Train a reward model, collect labels, and run PPO
8. Why should teams evaluate each task after a model merge?
   • Merging guarantees every benchmark improves
   • Evaluation is needed only to calculate the file size
   • A merged model cannot be run without retraining
   • Combining weight updates can create interference or reduce performance on a task