A small controlled comparison found that an adversary plus survival reward produced no reliable benefit over a matched non-competitive legal-reasoning training run.
Adversarial NLI: A New Benchmark for Natural Language Understanding
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Does the Competitive Component of Adversarial Self-Play Improve Legal Reasoning? A Controlled Negative Result
A small controlled comparison found that an adversary plus survival reward produced no reliable benefit over a matched non-competitive legal-reasoning training run.