princeton-nlp
commited on
Commit
•
eae0626
1
Parent(s):
641094a
Update README.md
Browse files
README.md
CHANGED
@@ -33,10 +33,15 @@ Contact: `{tianyug, awettig}@princeton.edu`
|
|
33 |
## Benchmarking results
|
34 |
|
35 |
|
36 |
-
|
37 |
|
38 |
![image/png](https://cdn-uploads.huggingface.co/production/uploads/607f846419a5af0183d7bfb9/78ppz9Y3-_fROfYIV_Re4.png)
|
39 |
|
|
|
|
|
|
|
|
|
|
|
40 |
You can find results for more tasks and models in this [spreadsheet](https://docs.google.com/spreadsheets/d/1qGzimBE8F896p1m7_yWHnjyGX7kpEAeyaT1h2iTbNzE/edit?usp=sharing). In this detailed results, we show that our model can retain the original Llama-3's general LM performance (on tasks selected by the [HF Open LLM Leaderboard v1](https://huggingface.co/spaces/open-llm-leaderboard-old/open_llm_leaderboard)). This is non-trivial in long-context fine-tuning and requires a careful selection of the fine-tuning data mixture and the training configurations.
|
41 |
|
42 |
|
|
|
33 |
## Benchmarking results
|
34 |
|
35 |
|
36 |
+
64K result:
|
37 |
|
38 |
![image/png](https://cdn-uploads.huggingface.co/production/uploads/607f846419a5af0183d7bfb9/78ppz9Y3-_fROfYIV_Re4.png)
|
39 |
|
40 |
+
512K result:
|
41 |
+
|
42 |
+
|
43 |
+
![image/png](https://cdn-uploads.huggingface.co/production/uploads/607f846419a5af0183d7bfb9/pwdlpXKSG68V0MxFKNOn1.png)
|
44 |
+
|
45 |
You can find results for more tasks and models in this [spreadsheet](https://docs.google.com/spreadsheets/d/1qGzimBE8F896p1m7_yWHnjyGX7kpEAeyaT1h2iTbNzE/edit?usp=sharing). In this detailed results, we show that our model can retain the original Llama-3's general LM performance (on tasks selected by the [HF Open LLM Leaderboard v1](https://huggingface.co/spaces/open-llm-leaderboard-old/open_llm_leaderboard)). This is non-trivial in long-context fine-tuning and requires a careful selection of the fine-tuning data mixture and the training configurations.
|
46 |
|
47 |
|