Update README.md
Browse files
README.md
CHANGED
@@ -63,6 +63,24 @@ Each instance consists of a question, a pair of contradictory passages extracted
|
|
63 |
|
64 |
The dataset has been used in the paper to assess LLMs performance when augmented with retrieved passages containing real-world knowledge conflicts.
|
65 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
66 |
### Out-of-Scope Use
|
67 |
|
68 |
<!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->
|
@@ -203,7 +221,7 @@ For further information, see the annotation guidelines of the paper.
|
|
203 |
|
204 |
<!-- This section describes the people or systems who created the annotations. -->
|
205 |
|
206 |
-
Yufang Hou, Alessandra Pascale, Javier Carnerero-Cano, Tigran Tchrakian, Radu Marinescu, Elizabeth Daly, Inkit Padhi
|
207 |
|
208 |
#### Personal and Sensitive Information
|
209 |
|
|
|
63 |
|
64 |
The dataset has been used in the paper to assess LLMs performance when augmented with retrieved passages containing real-world knowledge conflicts.
|
65 |
|
66 |
+
The following figure illustrates the evaluation process:
|
67 |
+
|
68 |
+
<p align="center">
|
69 |
+
<img src="./figs/Evaluation.png" width=70%/>
|
70 |
+
</p>
|
71 |
+
|
72 |
+
And the following table shows the performance of five LLMs (Mistral-7b-inst, Mixtral-8x7b-inst, Llama-2-70b-chat, Llama-3-70b-inst, and GPT-4) on the Wikipedia Contradict Benchmark based on rigorous human evaluations on a subset of answers for 55 instances, which corresponds to 1,375 LLM responses in total.
|
73 |
+
|
74 |
+
<p align="center">
|
75 |
+
<img src="./figs/table2.png" width=70%/>
|
76 |
+
</p>
|
77 |
+
|
78 |
+
Notes: “C”, “PC” and “IC” stand for “Correct”, “Partially correct”, “Incorrect”, respectively. “all”, “exp”, and “imp” represent for instance
|
79 |
+
types: all instances, instances with explicit conflicts, and instances with implicit conflicts. The
|
80 |
+
numbers represent the ratio of responses from each LLM that were assessed as “Correct, “Partially
|
81 |
+
correct, or “Incorrect for each instance type under a prompt template. The bold numbers highlight
|
82 |
+
the best models that correctly answer questions for each type and prompt template.
|
83 |
+
|
84 |
### Out-of-Scope Use
|
85 |
|
86 |
<!-- This section addresses misuse, malicious use, and uses that the dataset will not work well for. -->
|
|
|
221 |
|
222 |
<!-- This section describes the people or systems who created the annotations. -->
|
223 |
|
224 |
+
Yufang Hou, Alessandra Pascale, Javier Carnerero-Cano, Tigran Tchrakian, Radu Marinescu, Elizabeth Daly, Inkit Padhi
|
225 |
|
226 |
#### Personal and Sensitive Information
|
227 |
|