On-Policy Distillation Meets Off-Policy GRPO: Training Compact Instruction-Following Rerankers
Paper • 2609.01947 • Published
How to use from
llama.cppInstall from WinGet (Windows)
winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf everetteonhf/zerank-2-distilled-1B-GGUF:# Run inference directly in the terminal:
llama cli -hf everetteonhf/zerank-2-distilled-1B-GGUF:Use pre-built binary
# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf everetteonhf/zerank-2-distilled-1B-GGUF:# Run inference directly in the terminal:
./llama-cli -hf everetteonhf/zerank-2-distilled-1B-GGUF:Build from source code
git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf everetteonhf/zerank-2-distilled-1B-GGUF:# Run inference directly in the terminal:
./build/bin/llama-cli -hf everetteonhf/zerank-2-distilled-1B-GGUF:Use Docker
docker model run hf.co/everetteonhf/zerank-2-distilled-1B-GGUF:Quick Links
About
Static quants of anonymousauthor01/instruction_following_reranker (Prabhakar et al.) for use with llama.cpp.
Support for this model was added in PR #24083.
Files
| File | Format | Size |
|---|---|---|
zerank-2-distilled-1B.bf16.gguf |
BF16 | 2.31 GiB |
zerank-2-distilled-1B.Q6_K.gguf |
Q6_K | 974 MiB |
zerank-2-distilled-1B.Q4_K_M.gguf |
Q4_K_M | 770 MiB |
Original model card:
EMNLP 2026 Findings: On-Policy Distillation Meets Off-policy GRPO : Training Compact Instruction Following Rerankers
Model Checkpoint
Citation
If you use this model, please cite:
@inproceedings{prabhakar2026onpolicy,
title = {On-Policy Distillation Meets Off-Policy {GRPO}:
Training Compact Instruction-Following Rerankers},
author = {Prabhakar, Vignesh and Pan, Jialing and
Ankisettipalli, Anil Babu},
booktitle = {Findings of the Association for Computational Linguistics:
EMNLP 2026},
year = {2026},
address = {Budapest, Hungary},
publisher = {Association for Computational Linguistics},
note = {To appear}
}
- Downloads last month
- 272
Hardware compatibility
Log In to add your hardware
4-bit
6-bit
16-bit
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Install (macOS, Linux)
# Start a local OpenAI-compatible server with a web UI: llama serve -hf everetteonhf/zerank-2-distilled-1B-GGUF:# Run inference directly in the terminal: llama cli -hf everetteonhf/zerank-2-distilled-1B-GGUF: