hkust-nlp / deita

Deita: Data-Efficient Instruction Tuning for Alignment [ICLR2024]

hkust-nlp/deita Issues

Could you please publish the original data pool?
Closed 15 days ago2
The length of samples
Updated 15 days ago1
If I want to use vllm, which version should I install
Updated 2 months ago
Does the EVOL process of instruction dataset has been released?
Updated 2 months ago
reproduce mt-bench score
Updated 2 months ago1
Questions about performance improvement in Open LLM leaderboard
Updated 2 months ago3
Questions about the "Pool=50K" in your paper.
Closed 2 months ago2
[Question] Regarding the order bias in sample scoring.
Closed 2 months ago2
Question about which score to ultimately use for the filtering process.
Closed 2 months ago2
Have you conducted ablation experiments with three factors: Complexity, Quality, and Diversity? Which one has the greatest impact on performance improvement?
Updated 2 months ago1
Scorer models on hub are 7b not 13b
Closed 3 months ago2
Cosine distance computation
Closed 3 months ago4
How did you train the complexity & quality scorer
Closed 5 months ago8
Can we support more languages?
Updated 3 months ago1
[Question] Script to train Scorer model?
Closed 3 months ago1
What content is encoded when Llama13B encoded a sentence
Closed 4 months ago3
[question]Did you use the mean value of all token embedding in repr filter?
Closed 4 months ago3
request: is code of "Repr Filter" can be open source too?
Closed 5 months ago4
[question] repr_filter encoding only the instruction or instruction and anwser?
Closed 4 months ago1
[Question] Is the 6k dataset is a subset of 10k dataset.
Closed 4 months ago1
Some questions about running the scorer for arbitary model
Closed 5 months ago2
data of deita's dpo+sft
Closed 5 months ago