Tag
MaP-SQL enhances Text-to-SQL listwise selection without fine-tuning by employing structured memory retrieval and permutation-based ranking aggregation, outperforming state-of-the-art methods with reduced token usage.
The paper introduces GROUND, a governed semantic-retrieval framework that constrains LLM-generated analytics to approved business definitions, effectively reducing hallucinations in enterprise data analysis while ensuring compliance with security and data policies.
The article describes a reinforcement learning method with verifiable rewards to fine-tune AI models for text-to-SQL, achieving human-level accuracy on the BIRD benchmark without scaffolding.
This paper examines the trade-off between model size and inference compute for grammar-constrained text-to-SQL tasks using small language models, finding that beam search outperforms sample+vote and that scaling model size generally yields better results than scaling inference compute.
StocksTalk is a voice-enabled conversational agent that maps spoken natural language into executable SQL queries for financial data screening, integrating retrieval-augmented generation and human-in-the-loop verification to reduce errors.
The paper proposes TraceSQL, a lightweight and traceable verification model for text-to-SQL systems that uses explicit diagnostic features to estimate answerability without reference queries, achieving improved performance over existing baselines on the BIRD benchmark.
The paper proposes a structural abstention pattern for AI systems to prevent hallucinations in text-to-SQL applications by separating a trusted kernel for deterministic execution from a generative shell for interpretation, enhancing reliability in enterprise deployments.
This paper proposes SDAM, a memory-based framework for complex Text-to-SQL that uses structure-difference aware reasoning, contradiction-aware reflection, and schema-grounded memory evolution to improve SQL generation. Experiments show modest gains on BIRD-dev and Spider-test benchmarks.
This paper introduces the 'crystallization problem' for evaluating reusable memory in text-to-SQL systems, showing that storing verified corrected queries in a per-database bank improves held-out first-attempt accuracy by 4.34 points on BIRD, capturing 44.4% of the headroom provided by on-demand repair. Controlled interventions identify database-specific content as the main driver.
This paper introduces MERIT, a training-free agent that uses causal episodic memory of past repair outcomes to improve subsequent Text-to-SQL generations, boosting execution accuracy on Spider and BIRD benchmarks.
BAP-SQL presents a budget-aware observation planning approach for agentic text-to-SQL, treating SQL query choice as a budget-control decision to improve tight-budget success while reducing token usage.
AttnLink is a research paper presenting an attention-based framework for schema linking in Text-to-SQL, converting LLM internal attention into continuous relevance scores for schema items. Experiments on Spider, BIRD, and Spider2-SQLite show high mAP scores and millisecond-scale latency.
SERL-SQL proposes a selective execution-grounded reinforcement learning framework for multi-turn Text-to-SQL agents, using teacher-student likelihood gaps to reweight GRPO advantages on SQL action tokens. It achieves strong results on BIRD and Spider benchmarks.
This paper presents SAFAARI, a multi-agent framework that improves schema linking for NL-to-SQL systems in customer support, achieving an 81.66% SEAL score and 8x reduction in development time.
Proposes AutoThinkSQL, a framework that integrates an auto-thinking mechanism into SFT and DPO for Text-to-SQL, enabling the model to dynamically skip reasoning for simple queries and invoke deep CoT for complex ones, achieving gains on Spider and BIRD benchmarks while reducing output tokens by 24.6% and latency by 17.1%.
CHS-SQL proposes a confidence-guided heuristic search schema linking process for text-to-SQL using small language models, achieving state-of-the-art results by balancing precision and recall in schema selection.
EvoSQL is a co-evolution framework for Text-to-SQL that iteratively improves SQL generation via a generator-critic pair with episodic memory, achieving gains on Spider and BIRD benchmarks.
This article argues that text-to-SQL benchmarks must account for the complexities and challenges of real-world data stores, not just idealized datasets.
A weekly roundup of four trending open-source repos for AI agent builders, including FastMCP for Model Context Protocol, WrenAI for natural language database queries, KTransformers for LLM optimization, and Open Interpreter for a code-executing AI assistant.
This paper introduces PCC-SQL, a method for enforcing column-use policies in text-to-SQL generation by constrained decoding, achieving deterministic elimination of violations with 0% Leakage Rate and high coverage on benchmarks.