Selene 1 Mini: the best small language model-as-a-judge
•
12
Atla
Enterprise
company
Verified
AI & ML interests
Scalable oversight
Recent Activity
Articles
AtlaAI's activity
kaikaidai
updated
a
Space
19 days ago
kaikaidai
updated
a
model
19 days ago
kaikaidai
updated
a
Space
19 days ago
mbartolo
authored
a
paper
about 1 month ago
inwaves
authored
a
paper
3 months ago
spisupat
authored
a
paper
3 months ago
MauriceBurg
authored
a
paper
3 months ago
kaikaidai
authored
a
paper
3 months ago
NinaCalvi
authored
a
paper
3 months ago
Post
1081
📈 Early results on the 8B evaluation model we've been training...
@NinaCalvi wrote about the progress we've made this quarter towards training the best 'LLM-as-a-judge' evaluator. We've significantly improved against the baseline and are approaching state-of-the-art evaluation performance with an 8B model.
Next up: training Llama-3.1-70B 👀
Here's the full article: https://www.atla-ai.com/post/evaluating-the-evaluator
@NinaCalvi wrote about the progress we've made this quarter towards training the best 'LLM-as-a-judge' evaluator. We've significantly improved against the baseline and are approaching state-of-the-art evaluation performance with an 8B model.
Next up: training Llama-3.1-70B 👀
Here's the full article: https://www.atla-ai.com/post/evaluating-the-evaluator
- 2 replies
mbartolo
authored
10
papers
about 1 year ago
Beat the AI: Investigating Adversarial Human Annotation for Reading Comprehension
Paper
•
2002.00293
•
Published
Interpretation of Natural Language Rules in Conversational Machine Reading
Paper
•
1809.01494
•
Published
Winoground: Probing Vision and Language Models for Visio-Linguistic Compositionality
Paper
•
2204.03162
•
Published
•
1
Improving Question Answering Model Robustness with Synthetic Adversarial Data Generation
Paper
•
2104.08678
•
Published
Dynatask: A Framework for Creating Dynamic AI Benchmark Tasks
Paper
•
2204.01906
•
Published
Models in the Loop: Aiding Crowdworkers with Generative Annotation Assistants
Paper
•
2112.09062
•
Published
DMLR: Data-centric Machine Learning Research -- Past, Present and Future
Paper
•
2311.13028
•
Published
•
1
Human Feedback is not Gold Standard
Paper
•
2309.16349
•
Published
•
5
The PRISM Alignment Project: What Participatory, Representative and Individualised Human Feedback Reveals About the Subjective and Multicultural Alignment of Large Language Models
Paper
•
2404.16019
•
Published
•
1
Aya Dataset: An Open-Access Collection for Multilingual Instruction Tuning
Paper
•
2402.06619
•
Published
•
57