Step-2 SFT reference policies (π_ref) used to initialize DDRO (MS MARCO / NQ; PQ and Title+URL DocIDs); use these for fair comparisons/ablations.
Kidist Amde Mekonnen
kiyam
AI & ML interests
AI (Generative models, Computer vision, NLP) ,XAI
Recent Activity
updated a model 32 minutes ago
kiyam/ddro-nq-pq-sft updated a model about 7 hours ago
kiyam/ddro-msmarco-pq-sft authored a paper 7 days ago
Lightweight and Direct Document Relevance Optimization for Generative
Information RetrievalOrganizations
None yet
models 8
kiyam/ddro-nq-pq-sft
Text Generation • 0.3B • Updated • 5 • 1
kiyam/ddro-msmarco-pq-sft
Text Generation • 0.3B • Updated • 2 • 1
kiyam/ddro-nq-tu-sft
Text Generation • 0.2B • Updated • 5 • 1
kiyam/ddro-msmarco-tu-sft
Text Generation • Updated • 2 • 1
kiyam/ddro-nq-pq
0.3B • Updated • 2
kiyam/ddro-msmarco-pq
0.3B • Updated • 58
kiyam/ddro-nq-tu
0.3B • Updated • 2
kiyam/ddro-msmarco-tu
0.3B • Updated • 7 • 1