Instructions to use PengxinWang/RobustLLMAgent with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use PengxinWang/RobustLLMAgent with PEFT:
Task type is invalid.
- Notebooks
- Google Colab
- Kaggle
Remove redundant motivation README
Browse files
evaluation_results/analysis/motivation/README.md
DELETED
|
@@ -1,12 +0,0 @@
|
|
| 1 |
-
# Vanilla fragility
|
| 2 |
-
|
| 3 |
-
图示 step200 Vanilla 在 WebShop 和 ALFWorld 上的成功率。WebShop 为 evaluation seeds 0/1/2 的均值,每个 seed 500 goals;ALFWorld 为 seed0 的 ID 140 + OOD 134 个任务,按成功数和任务数合并。
|
| 4 |
-
|
| 5 |
-
不同环境使用不同扰动强度,按已有结果中较大退化选取,不能视为相同强度的跨环境比较。此图仅展示点估计,不含误差条。
|
| 6 |
-
|
| 7 |
-
`data/vanilla_fragility.csv` 为直接绘图数据;`webshop_source_rows.csv` 和 `alfworld_source_rows.csv` 保留逐 seed / split 的来源行与原始计数。
|
| 8 |
-
|
| 9 |
-
```bash
|
| 10 |
-
python -m pip install -r png/requirements.txt
|
| 11 |
-
python png/reproduce.py
|
| 12 |
-
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
evaluation_results/analysis/motivation/data/checksums.json
CHANGED
|
@@ -1,5 +1,4 @@
|
|
| 1 |
{
|
| 2 |
-
"README.md": "f90c474aca48f475472639e482d06f7effd4d1dbf1de68f9a795029a1b4302e6",
|
| 3 |
"data/alfworld_source_rows.csv": "d42b655fa82f93473a70481e707cc21180e030c6b77ded8f46d39e0472f8efda",
|
| 4 |
"data/provenance.json": "e81d2b3317c3f1dc7b8ccd2d07e0873004762415c865785ab63d7d0d358a3264",
|
| 5 |
"data/vanilla_fragility.csv": "41fad64390970ce481383a76eca93e1a1aa0c859d73449e58b417ea50b6d4e4c",
|
|
|
|
| 1 |
{
|
|
|
|
| 2 |
"data/alfworld_source_rows.csv": "d42b655fa82f93473a70481e707cc21180e030c6b77ded8f46d39e0472f8efda",
|
| 3 |
"data/provenance.json": "e81d2b3317c3f1dc7b8ccd2d07e0873004762415c865785ab63d7d0d358a3264",
|
| 4 |
"data/vanilla_fragility.csv": "41fad64390970ce481383a76eca93e1a1aa0c859d73449e58b417ea50b6d4e4c",
|