Reverse Probing: Evaluating Knowledge Transfer via Finetuned Task Embeddings for Coreference Resolution
This work addresses the problem of understanding task-specific representation adaptability for researchers in NLP, though it is incremental as it builds on classical probing methods.
The paper tackled evaluating knowledge transfer from simple source tasks to complex target tasks by using embeddings from tasks like paraphrase detection and named entity recognition for coreference resolution, finding that semantic similarity tasks were most beneficial and combining embeddings with attention-based aggregation improved performance.
In this work, we reimagine classical probing to evaluate knowledge transfer from simple source to more complex target tasks. Instead of probing frozen representations from a complex source task on diverse simple target probing tasks (as usually done in probing), we explore the effectiveness of embeddings from multiple simple source tasks on a single target task. We select coreference resolution, a linguistically complex problem requiring contextual understanding, as focus target task, and test the usefulness of embeddings from comparably simpler tasks tasks such as paraphrase detection, named entity recognition, and relation extraction. Through systematic experiments, we evaluate the impact of individual and combined task embeddings. Our findings reveal that task embeddings vary significantly in utility for coreference resolution, with semantic similarity tasks (e.g., paraphrase detection) proving most beneficial. Additionally, representations from intermediate layers of fine-tuned models often outperform those from final layers. Combining embeddings from multiple tasks consistently improves performance, with attention-based aggregation yielding substantial gains. These insights shed light on relationships between task-specific representations and their adaptability to complex downstream tasks, encouraging further exploration of embedding-level task transfer.