AI TRENDS | FLock Research Has 3 Papers Accepted to ECML PKDD 2026

FLock announced that three papers from its Research Team were accepted to ECML PKDD 2026, one of Europe’s major international conferences in machine learning, data mining, and knowledge discovery. According to Foresight News, the papers focus on common issues in large-model complex reasoning, including shortcut-taking, errors during reasoning, and gradual drift from the original goal in long reasoning chains. One study helps models maintain their original line of thought during long, multi-step reasoning, reducing deviation by 63% and lifting average accuracy by 7.4 percentage points across models of different sizes. The other two studies improve reasoning on unfamiliar problems and increase the efficiency of local correction after reasoning errors are detected. FLock Research said it continues basic research on large-model training, reasoning, and reliability, aiming to improve accuracy, stability, and generalization in complex tasks.