Augustus Odena commited on
Commit
5d6d122
1 Parent(s): 73445a4

fix readme

Browse files
Files changed (1) hide show
  1. README.md +0 -2
README.md CHANGED
@@ -29,8 +29,6 @@ The model can use its existing position embeddings to reason about different ima
29
  - **Resources for more information:** Check out our [blog post](https://www.adept.ai/blog/fuyu-8b).
30
 
31
  ## Evaluation
32
- The chart above evaluates user preference for SDXL (with and without refinement) over SDXL 0.9 and Stable Diffusion 1.5 and 2.1.
33
- The SDXL base model performs significantly better than the previous variants, and the model combined with the refinement module achieves the best overall performance.
34
  Though not the focus of this model, we did evaluate it on standard image understanding benchmarks:
35
 
36
  | Eval Task | Fuyu-8B | Fuyu-Medium | LLaVA 1.5 (13.5B) | QWEN-VL (10B) | PALI-X (55B) | PALM-e-12B | PALM-e-562B |
 
29
  - **Resources for more information:** Check out our [blog post](https://www.adept.ai/blog/fuyu-8b).
30
 
31
  ## Evaluation
 
 
32
  Though not the focus of this model, we did evaluate it on standard image understanding benchmarks:
33
 
34
  | Eval Task | Fuyu-8B | Fuyu-Medium | LLaVA 1.5 (13.5B) | QWEN-VL (10B) | PALI-X (55B) | PALM-e-12B | PALM-e-562B |