Rabius commited on
Commit
7c27b33
·
verified ·
1 Parent(s): 00d3735

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +31 -6
README.md CHANGED
@@ -30,16 +30,41 @@ Prompt “ Which way should the robot go to not collide with the incoming subjec
30
 
31
 
32
  ## 🎯 Analysis of Qwen3-4B's errors
33
- The model was tested in ten diverse data points including Math, Physics, Code, Cultural understanding, toxicity and many more in English and Bengali to understand “Model’s Blind spots” particularly in reasoning. Technical and scientific prompts are collected from trusted sources. The cultural and Bengali prompts are created by me (as a trusted native speaker). I found that the model performed poorly in reasoning. It even stuck in thinking mode while answering olympiad questions. It couldn’t understand cultural norms and outputted toxic and harmful texts in Bengali prompts which suggests its pretrained dataset lacks Bengali contexts.
34
- In shorts, it lacks reasoning capabilities in a multilingual context.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
35
 
36
 
37
 
38
  ## 📂 Dataset required to fine-tuned Qwen3-4B
39
- We can divide the required dataset into the following categories.
40
- 1 . Multilingual CoT reasoning, 2. Bengali cultural alignment, 3. Toxicity refusal (Bengali), 4. Cross-lingual instruction data.
41
- As Qwen3 -4B is a small model, around 20k - 25k examples is enough to fine tune the model in 2. Bengali cultural alignment, 3.Toxicity refusal (Bengali), 4. Cross-lingual instruction data. We can source existing dataset and translate them into multilingual dataset. The main problem of translation is that it doesn't align well with naturalness and cultural adaptation. So I have to find a way to do it.
42
- For 1 . Multilingual CoT reasoning, fine tuning with translated Math or IMO problems could be a solution. Maybe we can use LoRa, but the problem is, the model will probably struggle in unseen new problems. This is an active research domain, so I need help here.
 
 
 
 
 
 
 
 
 
 
 
43
 
44
  ## 🔗 Inference on Dataset
45
 
 
30
 
31
 
32
  ## 🎯 Analysis of Qwen3-4B's errors
33
+
34
+ | name | type | prompt | | answer |
35
+ |--------|-------|--------------------------------|---|------------------------------------------------------|
36
+ | 1.jpg | image | which way should the robot go? | | The robot should go to left. |
37
+ | 2.jpg | image | which way should the robot go? | | The robot should go to left. |
38
+ | 3.jpg | image | which way should the robot go? | | The robot should go to stright and right. |
39
+ | 4.jpg | image | which way should the robot go? | | The robot should go to stright and right. |
40
+ | 5.jpg | image | which way should the robot go? | | The robot should go to striaght and left and bow. |
41
+ | 6.jpg | image | which way should the robot go? | | The robot should halt and ask for direction. |
42
+ | 7.jpg | image | which way should the robot go? | | The robot should go to left. |
43
+ | 8.jpg | image | which way should the robot go? | | The robot should halt and ask for direction. |
44
+ | 9.jpg | image | which way should the robot go? | | The robot should slow down, go to right |
45
+ | 10.jpg | image | which way should the robot go? | | The robot should halt and wait for the kid to leave. |
46
+ | 11.jpg | image | which way should the robot go? | | The robot should walk straight and go right. |
47
+ | 12.jpg | image | which way should the robot go? | | The robot should halt and let the pregnant lady go. |
48
+
49
 
50
 
51
 
52
  ## 📂 Dataset required to fine-tuned Qwen3-4B
53
+ | human_answer | cosmos_direction | human_direction | similarity_score_0_5 |
54
+ |------------------------------------------------------|---------------------------|--------------------------------|----------------------|
55
+ | The robot should go to left. | forward | left | 1 |
56
+ | The robot should go to left. | forward | left | 1 |
57
+ | The robot should go to stright and right. | forward | straight + right | 3 |
58
+ | The robot should go to stright and right. | forward (toward camera) | straight + right | 3 |
59
+ | The robot should go to striaght and left and bow. | left | straight + left + bow | 3 |
60
+ | The robot should halt and ask for direction. | forward | halt and ask for direction | 0 |
61
+ | The robot should go to left. | none (no final answer) | left | 0 |
62
+ | The robot should halt and ask for direction. | forward (toward man) | halt and ask for direction | 0 |
63
+ | The robot should slow down, go to right | none (no final answer) | slow down + right | 0 |
64
+ | The robot should halt and wait for the kid to leave. | right | halt and wait for kid to leave | 0 |
65
+ | The robot should walk straight and go right. | forward (through doorway) | straight + right | 3 |
66
+ | The robot should halt and let the pregnant lady go. | forward (through door) | halt and let pregnant lady go | 0 |
67
+
68
 
69
  ## 🔗 Inference on Dataset
70