arXiv subjects

Qunhui Zhang

Publications and source records attributed to Qunhui Zhang.

3 recordsLinked to original sources

arXiv2026

Generic Constraints Projection: Four-Dimensional Type Inference for Dynamic Languages

Type inference for dynamically typed languages must reconcile four qualitatively different sources of evidence: assigned values, explicit declarations, contextual requirements, and structural operations. Existing approaches often combine them into one constraint set, causing spurious conflicts or requiring annotations. We present Generic Constraints Projection (GCP), a zero-annotation inference framework that stores these sources in four monotone slots on a stable definition-time template and evaluates each call in a fresh projection session, preventing cross-call contamination while specializing return types. GCP uses Outline Equational Matching, an open structural preorder, and a future-this projection rule that preserves concrete receiver types across fluent chains and subtype extensions. On the success-state fragment of a bounded type domain, we prove monotonicity, local and global fixed-point convergence, conditional projection soundness, termination, multi-module convergence, and order independence. For an immutable core language, we also prove big-step evaluation existence, type preservation, runtime receiver retention, and projection-evaluation coherence. We instantiate GCP in Outline for typed ontology worlds and in a Python annotation-recovery pipeline. On 513 manually adapted, fact-paired Outline ports of TypeEvalPy cases, GCP obtains 513/513 exact matches, compared with 485/513 for the published Codestral Q&A baseline on the same fact IDs (two-sided exact McNemar p = 7.45e-9). This is a carrier-port evaluation in TypeEvalPy's closed-world Python vocabulary, not a run on unmodified Python sources.

cs.PL↗

arXiv2026

AiFlow: Token-Native Reactive Orchestration with Bounded Backpressure for Streaming LLM Applications

Large language model (LLM) applications increasingly operate as streaming workflows combining retrieval, tool calls, safety filters, and multi-agent coordination. Although contemporary frameworks expose provider deltas, workflow nodes often treat generation as coarse request-response steps, leaving queue management, worker allocation, ordering, and backpressure to ad hoc callback code. This paper presents AiFlow, a token-native reactive orchestration model that normalizes provider deltas into typed Context events propagated through a directed streaming graph. Each node is managed by a Node Guardian that declares and enforces local queue bounds, worker concurrency, ordering, overflow policy, cancellation propagation, and retry discipline. We formalize the bounded-memory property, present the compilation from a compact DSL and JSON graph form, and provide static validation for type safety, state concurrency, and injection compatibility. Controlled microbenchmarks, captured DeepSeek trace replay (30 runs), descriptive online runs, LangGraph baselines, a streaming RAG workload, and an Ollama local-backend check show that AiFlow does not alter provider-side Model TTFT but reduces Application TTFPT by 70.9-94.7\% versus aggregation and keeps runtime-owned queue depth within declared bounds (93.7-96.5\% MaxQ reduction versus unbounded policies). The supplementary artifact contains scripts, raw traces, machine-readable tables, checksums, and an API-free smoke test; the public implementation is available through the FIT Framework repository.

cs.SE↗

arXiv2026

VirtualSet: Typed Ontology Worlds as an LLM Generation Target for Grounded Queries and Guarded Decisions

Large language models increasingly read and act on enterprise data, but SQL gives a late error signal: hallucinated fields or relations can execute and return plausible wrong answers, while incorrect writes cannot be safely assessed after execution. We present VirtualSet, a live, receiver-typed ontology-world interface and generation target for LLMs. Instead of SQL, the model emits set expressions over entity-edge worlds. Generic Constraint Projection (GCP) checks expressions before execution, while future this preserves concrete receiver types through collection chains, turning invalid fields, edges, receivers, and actions into token-anchored type errors. Type-clean reads use a SQL fast path or bounded stream interpretation, with a parity oracle checking both paths over the exercised operator space. The same substrate supports guarded decisions: actions run first in a simulated world, and world-change events require external approval before actualization. On BIRD, we lift relational schemas into typed worlds and compare VirtualSet with direct SQL while holding the model, evidence, values, zero-shot setting, timeout, glossary, repair/voting, and grader constant where possible. On a frozen 1,072-question split, VirtualSet achieves 67.5% accuracy versus 63.5% for glossary-matched direct SQL with repair and voting (+4.0 points; McNemar exact p = 0.00117) using deepseek-reasoner. Full-corpus analysis finds no engine mis-computation of a type-clean expression; remaining errors arise from model semantics or gold defects. In a 30-body guard corpus, the write chain intercepts 20/20 hallucinated action bodies with zero false positives. VirtualSet thus remains competitive on SQL's home benchmark while providing pre-execution semantics for guarded decisions.

cs.PL↗