Report for ProsusAI/finbert
#85
by
giskard-bot
- opened
Hi Team,
This is a report from Giskard Bot Scan 🐢.
We have identified 3 potential vulnerabilities in your model based on an automated scan.
This automated analysis evaluated the model on the dataset financial_phrasebank (subset sentences_allagree, split train).
👉Performance issues (2)
| Vulnerability | Level | Data slice | Metric | Transformation | Deviation |
|---|---|---|---|---|---|
| Performance | medium 🟡 | avg_whitespace(text) < 0.160 AND avg_whitespace(text) >= 0.156 |
Precision = 0.917 | — | -5.67% than global |
🔍✨Examples
For records in the dataset where `avg_whitespace(text)` < 0.160 AND `avg_whitespace(text)` >= 0.156, the Precision is 5.67% lower than the global Precision.| text | avg_whitespace(text) | label | Predicted label |
|
|---|---|---|---|---|
| 533 | According to Finnair Technical Services , the measure is above all due to the employment situation . | 0.16 | neutral | positive (p = 0.50) |
| 841 | Previously , EB delivered a custom solution for LG Electronics and now is making it commercially available for other mobile terminal vendors as well as to wireless operators . | 0.16 | positive | neutral (p = 0.76) |
| 1062 | The contract covers turnkey deliveries to all five airports operated by the authority -- John F Kennedy , LaGuardia , Newark , Teterboro and Stewart International . | 0.158537 | neutral | positive (p = 0.70) |
| Vulnerability | Level | Data slice | Metric | Transformation | Deviation |
|---|---|---|---|---|---|
| Performance | medium 🟡 | text_length(text) >= 99.500 AND text_length(text) < 106.500 |
Precision = 0.921 | — | -5.22% than global |
🔍✨Examples
For records in the dataset where `text_length(text)` >= 99.500 AND `text_length(text)` < 106.500, the Precision is 5.22% lower than the global Precision.| text | text_length(text) | label | Predicted label |
|
|---|---|---|---|---|
| 533 | According to Finnair Technical Services , the measure is above all due to the employment situation . | 100 | neutral | positive (p = 0.50) |
| 1413 | The contract also includes installation work in a new multistorey carpark for close on 1,000 vehicles . | 103 | neutral | positive (p = 0.55) |
| 1496 | The repo rate will gradually reach 2 % at the end of 2010 , according to Nordea 's Economic Outlook . | 101 | neutral | positive (p = 0.51) |
👉Robustness issues (1)
| Vulnerability | Level | Data slice | Metric | Transformation | Deviation |
|---|---|---|---|---|---|
| Robustness | medium 🟡 | — | Fail rate = 0.062 | Add typos | 62/1000 tested samples (6.2%) changed prediction after perturbation |
🔍✨Examples
When feature “text” is perturbed with the transformation “Add typos”, the model changes its prediction in 6.2% of the cases. We expected the predictions not to be affected by this transformation.| text | Add typos(text) | Original prediction | Prediction after perturbation | |
|---|---|---|---|---|
| 1919 | Cash flow from operations totalled EUR 7.4 mn , compared to a negative EUR 68.6 mn in the second quarter of 2008 . | Cash dlow from operations totalled WUR 7.4 mhn , comlared to a negative EUR 68.6 mn in the second wquarter of 2008 . | positive (p = 0.86) | negative (p = 0.95) |
| 1584 | `` These developments partly reflect the government 's higher activity in the field of dividend policy . '' | `` These developmsents partly reflect he government 's higher activity in the field of dividend policty . '' | positive (p = 0.75) | neutral (p = 0.58) |
| 628 | The share of the share capital of both above mentioned shareholders remains below 5 % . | The share of the share capiral of both above mentioned sharehooders remains below 5 % . | neutral (p = 0.72) | negative (p = 0.83) |
Checkout out the Giskard Space and test your model.
Disclaimer: it's important to note that automated scans may produce false positives or miss certain vulnerabilities. We encourage you to review the findings and assess the impact accordingly.