English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Large Language Models (LLMs) can justify or criticize their predictionsthrough discussion with other models or humans, thereby enhancing theirintrinsic understanding of instances. While proactive discussions enhanceperformance, this approach is currently limited to the inference phase. In thiscontext, we posit a hypothesis: learning interactive discussions duringtraining can improve understanding for the instances in the training step andproficiency in logical/critical thinking ability and verbalized expression ofthe model in the inference step. Our proposed SAIE training method involvesboth supportive and adversarial discussions between the learner and partnermodels. The learner model receives a remark from the partner through thediscussion, and the parameters of the learner model are then updated based onthis remark. That is, the teacher signal dynamically adjusts in response to theevolving model output throughout the training step. By bolstering the capacityfor discussion and comprehension of instances, our experiments across datasets,including GSM8K, CommonsenseQA, and MMLU, reveal that models fine-tuned withour method consistently surpass those trained with standard fine-tuningtechniques. Moreover, our approach demonstrates superior performance inmulti-agent inference scenarios, boosting the models' reasoning abilities atthe inference step.

SAIE Framework: Support Alone Isn't Enough -- Advancing LLM Training with Adversarial Remarks