$A^3$-Bench: Benchmarking Memory-Driven Scientific Reasoning via Anchor and Attractor Activation

TL;DR

This paper introduces $A^3$-Bench, a new benchmark for evaluating scientific reasoning that emphasizes memory-driven mechanisms like anchor and attractor activation, filling a gap in existing evaluation methods.

Contribution

It proposes a novel benchmark and evaluation framework focusing on memory activation in scientific reasoning, including annotations, a dual-scale memory evaluation, and a new metric.

Findings

Memory activation correlates with reasoning performance.

The benchmark enables analysis of memory-driven reasoning mechanisms.

Different models show varied memory utilization patterns.

Abstract

Scientific reasoning relies not only on logical inference but also on activating prior knowledge and experiential structures. Memory can efficiently reuse knowledge and enhance reasoning consistency and stability. However, existing benchmarks mainly evaluate final answers or step-by-step coherence, overlooking the \textit{memory-driven} mechanisms that underlie human reasoning, which involves activating anchors and attractors, then integrating them into multi-step inference. To address this gap, we propose $A^3$-Bench~ https://a3-bench.github.io, a benchmark designed to evaluate scientific reasoning through dual-scale memory-driven activation, grounded in Anchor and Attractor Activation. First, we annotate 2,198 science reasoning problems across domains using the SAPM process(subject, anchor & attractor, problem, and memory developing). Second, we introduce a dual-scale memory evaluation framework utilizing anchors and attractors, along with the AAUI(Anchor--Attractor Utilization Index) metric to measure memory activation rates. Finally, through experiments with various base models and paradigms, we validate $A^3$-Bench and analyze how memory activation impacts reasoning performance, providing insights into memory-driven scientific reasoning.