# Beyond Accuracy: Introducing a Symbolic-Mechanistic Approach to Interpretable Evaluation

arXiv: https://arxiv.org/abs/2603.23517 · HTML: https://mlsanity.com/paper/2603.23517/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Builds on 7 of its 15 references that are in the map.
- First posted on arXiv 6 Mar 2026.

## Builds on (most cited first, 7 of 7)

- [Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference](https://mlsanity.com/paper/1902.01007/) (2019)
- [Progress measures for grokking via mechanistic interpretability](https://mlsanity.com/paper/2301.05217/) (2023)
- [Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets](https://mlsanity.com/paper/2201.02177/) (2022)
- [TinyStories: How Small Can Language Models Be and Still Speak Coherent English?](https://mlsanity.com/paper/2305.07759/) (2023)
- [Is LLM-as-a-Judge Robust? Investigating Universal Adversarial Attacks on Zero-shot LLM Assessment](https://mlsanity.com/paper/2402.14016/) (2024)
- [Towards Data Contamination Detection for Modern Large Language Models: Limitations, Inconsistencies, and Oracle Challenges](https://mlsanity.com/paper/2409.09927/) (2024)
- [Benchmark Profiling: Mechanistic Diagnosis of LLM Benchmarks](https://mlsanity.com/paper/2510.01232/) (2025)

## Built on by (most cited first, 1 of 1)

- [Symbolic Mechanistic Data Attribution: Tracing Training Influence to Learned Behavioral Policies](https://mlsanity.com/paper/2606.29171/) (2026)

## Abstract (arXiv)

Accuracy-based evaluation cannot reliably distinguish genuine generalization from shortcuts like memorization, leakage, or brittle heuristics, especially in small-data regimes. In this position paper, we argue for mechanism-aware evaluation that combines task-relevant symbolic rules with mechanistic interpretability, yielding algorithmic pass/fail scores that show exactly where models generalize versus exploit patterns. We demonstrate this on NL-to-SQL by training two identical architectures under different conditions: one without schema information (forcing memorization), one with schema (enabling grounding). Standard evaluation shows the memorization model achieves 94% field-name accuracy on unseen data, falsely suggesting competence. Our symbolic-mechanistic evaluation reveals this model violates core schema generalization rules, a failure invisible to accuracy metrics.