tutor-robustness-eval/script/base_adversarial_agent.sh at main · epfl-ml4ed/tutor-robustness-eval · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
# !/bin/bash

# base_student_adversarial_agent vs base_incontext_tutor
python evaluation_parallel.py \
          --config-name parallel_evaluation \
          evaluation.strategies="['direct_request','emotional_threat','intentional_wrong_answer']" \
          attacker.type="base_student_adversarial_agent" \
          tutor.type="base_incontext_tutor" \
          agent.tutor.model_name="Qwen/Qwen2.5-7B-Instruct" \
          agent.tutor.vllm_port=8000 \
          agent.tutor.base_url="http://localhost:8000/v1" \
          agent.tutor.prompt_file="src/configs/prompt/tutor/default.txt" \
          agent.student.prompt_file="src/configs/prompt/student/default_6.txt" \
          agent.student.response_format="AdversarialPromptWithoutReason" \
          agent.tutor.response_format=null \
          wandb.name="base_student_adversarial_agent__base_incontext_tutor"


# base_student_adversarial_agent vs tutor with reasoning
python evaluation_parallel.py \
          --config-name parallel_evaluation \
          evaluation.strategies="['direct_request','emotional_threat','intentional_wrong_answer']" \
          attacker.type="base_student_adversarial_agent" \
          tutor.type="tutor_with_reasoning" \
          agent.tutor.model_name="Qwen/Qwen2.5-7B-Instruct" \
          agent.tutor.vllm_port=8000 \
          agent.tutor.base_url="http://localhost:8000/v1" \
          agent.tutor.prompt_file="src/configs/prompt/tutor/default_with_reason.txt" \
          agent.student.prompt_file="src/configs/prompt/student/default.txt" \
          agent.student.response_format="AdversarialPromptWithoutReason" \
          agent.tutor.response_format=GuidancePrompt \
          wandb.name="base_student_adversarial_agent__tutor_with_reasoning"


# base_student_adversarial_agent vs multi_agent_tutor
python evaluation_parallel.py \
          --config-name parallel_evaluation \
          evaluation.strategies="['direct_request','emotional_threat','intentional_wrong_answer']" \
          attacker.type="base_student_adversarial_agent" \
          tutor.type="multi_agent_tutor" \
          agent.tutor.model_name="Qwen/Qwen2.5-7B-Instruct" \
          agent.tutor.vllm_port=8000 \
          agent.tutor.base_url="http://localhost:8000/v1" \
          agent.tutor.prompt_file="src/configs/prompt/tutor/default.txt" \
          agent.student.prompt_file="src/configs/prompt/student/default.txt" \
          agent.student.response_format="AdversarialPromptWithoutReason" \
          agent.tutor.response_format=null \
          agent.reflector.tutor_model_name="Qwen/Qwen2.5-7B-Instruct" \
          agent.reflector.tutor_vllm_port=8000 \
          agent.reflector.tutor_base_url="http://localhost:8000/v1" \
          playground.tutor_reflection=true \
          wandb.name="base_student_adversarial_agent__multi_agent_tutor"