We lifted 4 functions out of this paper's own repositories and ran 4 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.
| Repository | Role | Ran |
|---|---|---|
| trusthlt/dp-across-nlp-tasks | canonical | 4 of 4 |
| Function | Status | Where it lives |
|---|---|---|
| categorical_accuracy | Ran | trusthlt/dp-across-nlp-tasks/NLPCode/natural_language_inference/utils.py code served (permissive licence) · get_code("c8e199b09d57c824") |
| epoch_time | Ran | trusthlt/dp-across-nlp-tasks/NLPCode/named_entity_recognition/utils.py code served (permissive licence) · get_code("3e21e6589663b136") |
| get_acc_pre_rec_f1 | Ran | trusthlt/dp-across-nlp-tasks/NLPCode/named_entity_recognition/utils.py code served (permissive licence) · get_code("b6e673e2fb74169b") |
| take_no_pad | Ran | trusthlt/dp-across-nlp-tasks/NLPCode/named_entity_recognition/utils.py code served (permissive licence) · get_code("593b63e88677631d") |
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
Preserving privacy in contemporary NLP models allows us to work with sensitive data, but unfortunately comes at a price. We know that stricter privacy guarantees in differentially-private stochastic gradient descent (DP-SGD) generally degrade model performance. However, previous research on the efficiency of DP-SGD in NLP is inconclusive or even counter-intuitive. In this short paper, we provide an extensive analysis of different privacy preserving strategies on seven downstream datasets in five different `typical' NLP tasks with varying complexity using modern neural models based on BERT and XtremeDistil architectures. We show that unlike standard non-private approaches to solving NLP tasks, where bigger is usually better, privacy-preserving strategies do not exhibit a winning pattern, and each task and privacy regime requires a special treatment to achieve adequate performance.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2112.08159")
get_code_for_paper("2112.08159")
have("2112.08159")
Connect an agent — have() is free.