QBX: A Compiler for Well-Formed, Auditable Alpha Programs
AuthorsYichong Bai, Kaigui Bian
AffiliationsThe Hong Kong Polytechnic University, Hong Kong, China · Peking University, Beijing, China
QBX turns financial alpha strategies into auditable programs that are easier to validate, reproduce, and search efficiently.
Key results
Versioned blocks define the compiler’s current strategy vocabulary.
Compared with 26.8% for free strings and 36.4% for a grammar-only control.
Times faster than full re-evaluation, with numerical parity.
Lower end of the 0.036–0.038 improvement over LLM-only books at matched budgets.
What the paper found
QBX is a compiler-based substrate for building and auditing formulaic alpha strategies. Instead of treating candidate formulas as free text, it admits whole strategies as typed port-DAGs drawn from a versioned registry of 95 blocks. Its compiler checks graph structure, types, executable linkage, and time-availability effects that reject contract-visible look-ahead, then emits an evidence-carrying representation with Merkle-style identities for reuse and trial accounting. On Chinese A-share panels, evaluation-valid candidates rose from 26.8% for free strings and 36.4% for a grammar-only control to 97.7% for typed programs; the checker found zero registered-contract violations across 23342 candidates. Incremental replay was 7.50× faster, with no value mismatches in 1760 comparisons. At matched LLM-call budgets, a hybrid of LLM-sourced breadth and compiler-closed search depth improved test RankICIR by 0.036–0.038. Experiments included a frozen DeepSeek-chat proposal pool, while the paper notes that it has no matched performance benchmark against Qlib. QBX improves validity, reuse, and auditability within its registered vocabulary; it does not add market information or establish that the restricted vocabulary captures every useful alpha.
Original abstract
Formulaic alpha miners increasingly differ in how they generate candidates, yet each system typically couples its own representation, evaluator, and selector into a bespoke pipeline whose results are difficult to reproduce or compare. We present QBX, a compiler that defines a well-formed, auditable space of alpha programs. A finite, versioned block registry and whole-strategy port-DAG language define the candidate space; the compiler either rejects a proposal with structured diagnostics or emits an evidence-carrying IR with types, time-availability effects, executable linkage, and Merkle-style content identities. Compiler-closed actions keep search inside the admitted space, node identities support incremental evaluation, and context-bound root identities govern program deduplication and trial provenance. The availability-effect system rejects contract-visible look-ahead while admitting legal point-in-time forecasts. On Chinese A-share panels, evaluation-valid proposals increase from 26.8% for free strings and 36.4% for a grammar-only control to 97.7% for typed programs. Incremental evaluation replays search lineages 7.5 times faster with numerical parity, while the availability checker admits zero registered-contract violations across 23,342 candidates. Under matched LLM-call budgets, compiler-closed search reaches the pool-specific diagnostic frontier with substantially fewer calls. These results concern validity, reuse, and search efficiency within the registered vocabulary, not an increase in the information available from the market panel.
Read the original paperMore in Efficient AI
Browse all 79 papers →Latent Feedback Is a Compute and Data Amplifier
Benhao Huang, Chaitanya Dwivedi, Xuezhe Ma, Alexander Moreno, Hector Liu
Latent feedback may let language models get more value from each training token and FLOP than simply repeating the data.
MACaQ: Fast, Model Agnostic Calibrated Quantization for Low-Precision Weights and Activations
Lorenz K. Muller, Philippe Bich, Chiara Boretti, Hyun-Min Chang, Jiawei Zhuang, Lukas Cavigelli
MACaQ aims to make large neural networks cheaper to run by quantizing their weights and activations quickly, with little added inference overhead.
Mobile AI Factory: Building Specialized Language Models for Your Mobile Device
Evan Li, Krishna Deshpande, Wen Cheng, Shyamnath Gollakota
Mobile AI Factory automatically builds and tunes a small language model for a specific task and phone, aiming to deliver strong performance with less compute.