Researchers build neurosymbolic router that sends math to exact solvers on Raspberry Pi
A paper on arXiv describes a neurosymbolic router that classifies incoming queries and routes structured tasks such as arithmetic and logic to deterministic solvers, while sending open-ended word problems to a small language model. The router is a learned deterministic finite automaton trained with the L grammatical inference algorithm, using the SLM as a membership oracle and labeled data as an…
Key points
- Router learns a DFA with L* algorithm to dispatch queries to exact solvers or SLM
- 98.3% overall accuracy on 100 test prompts, versus 72% for Program-of-Thought baseline
- Runs 8.8x faster and 2.8x more energy-efficient on Raspberry Pi 4B (8 GB, no GPU)
On a Raspberry Pi 4B with 8 GB RAM and no GPU, the system was tested on 100 unseen prompts from DeepMind Mathematics, GSM8K, and RuleTaker. It achieved 100% routing accuracy and 98.3% overall accuracy with a 512-token budget, compared with 72.0% for a Program-of-Thought baseline and 58.7% for a tool-calling agent. Formatted queries are answered in 1-11 ms, and a 30-token configuration runs 8.8x faster and 2.8x more energy-efficiently than Program-of-Thought.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Mirror-Score benchmarks D-peptide design tools against real-world affinity · 1 src
- Researchers introduce coffee framework for discrete diffusion model guidance · 2 src
- Researchers propose NashEval for context-dependent AI agent evaluation · 2 src
- Researchers test whether AI harnesses specialize or just repeat answers · 1 src
- Study finds synthetic embeddings match text for LLM fine-tuning · 1 src
Comments
via GitHub Discussions