← BEA 2026 paper

Reasoning chains: Qwen3.5-9B (think mode)

Qwen3.5-9B in think mode, evaluated on all 612 LoRuGEC test items. Each row shows the source sentence (with edited tokens highlighted), the model's prediction, and is expandable to read the full reasoning chain. Filter by rule, search the reasoning text, or jump straight to the JSONL.