Search arXiv⌕ Search

arXiv subjects

Nan Tang

Publications and source records attributed to Nan Tang.

At least 19 recordsLinked to original sources
arXiv2026

Post-SQLFix: Property-Guided Repair of LLM-Generated SQL with Runtime Verification

Large Language Models (LLMs) for Text-to-SQL frequently generate queries with syntax errors, invalid identifier references, or violations of SQL scoping and typing rules. Repairing these errors is a program correctness problem, yet existing approaches rely on Execution Feedback (EF), whose symptom-level, dialect-specific error messages force the LLM to speculate on root causes. We identify an Arbitration Paradox in this paradigm: the system relies on unverified models both to propose repairs and to make the final judgment on whether those repairs satisfy the required properties. We resolve this paradox by reformulating Text-to-SQL repair as property-guided program repair. Our framework, Post-SQLFix, canonicalizes SQL into a dialect-aware Canonical Query Structure (CQS) and validates it through a hierarchical static analyzer with three stratified property layers (Syntactic, Context-Free Structural, Context-Sensitive Semantic). The analyzer produces precise, schema-aware violation diagnostics to guide the LLM's repair proposals. A Verify-Before-Apply Temporal Monitor then accepts only repairs that reduce the violation count, do not reintroduce resolved violations, and satisfy the operator contracts. These runtime invariants are bounded by the declared rule family and operator contracts, while a finite retry budget ensures halting. On three established benchmarks (BIRD, Spider, Mini-Dev) spanning over 3,000 evaluation instances, three SQL dialects (SQLite, MySQL, PostgreSQL), and six model configurations, Post-SQLFix resolves 77.8% of the execution errors that EF leaves unrepaired, while requiring less than half as many repair iterations and 30% fewer tokens. Crucially, accepted transitions under the monitor introduce no statically detectable structural regressions in our evaluation, translating checked repair fidelity into absolute accuracy improvements of up to +11.6 percentage points.

cs.SE↗
arXiv2026

Back to the Future: The Evolving Universal Relation Model

Document backed analytics need databases that can be built as queries arrive. Enterprises increasingly ask analytical queries over messy document collections such as contracts, invoices, reports, emails, policies, and tickets, where the relevant structure and facts are often not known in advance. RAG can retrieve and summarize supporting evidence, but it does not by itself produce reusable, joinable, governed database state. Schema first construction offers the opposite strength: once a database is built, it supports joins, aggregation, constraints, indexes, optimization, and transactions. Yet for document analytics, fixing the schema and extracting all possible fields upfront is both expensive and brittle, because later queries and new documents often surface concepts that were not modeled before. We propose the Evolving Universal Relation Model (EUR), a query first model for document backed analytics. Users ask natural language queries as if the document corpus exposed one conceptual relation over application concepts. Internally, the DBMS turns each query into controlled database evolution: it updates the logical schema when new concepts are needed, extracts and integrates the facts required to answer the query, and decides what durable database state should remain for future queries. Thus, in EUR, a query is not only evaluated; it also helps construct the database. Document backed analytics becomes query first universal access plus workload-driven database schema and instance evolution.

cs.DB↗
arXiv2026

DocPilot: Interactive Agent Orchestration and Training for Long Visual Document Understanding

Multimodal long-document question answering requires models to seek sparse evidence across pages and modalities, understand heterogeneous document content, and reason over the collected evidence. Even advanced vision-language models (VLMs) can overlook sparse evidence and struggle with cross-modal reasoning in a single pass. Existing document agents further introduce multi-step interaction, but some still rely on predefined workflows or coarse-grained actions, limiting the flexibility and granularity of document reasoning. We introduce DocPilot, an agentic document framework that supports expressive orchestration and makes complex document solving learnable by small language models (SLMs). DocPilot orchestrates evidence seeking through page-level navigation and fine-grained evidence localization, and integrates local visual inspection, code-based numerical reasoning, and cross-modal reasoning for multimodal understanding. To transfer these capabilities to small models, we propose curriculum-based agentic training that progresses from single-ability learning to adaptive orchestration through supervision from data-grounded reasoning traces and executable interaction trajectories. Across MMLongBench-Doc, DocBench, FinRAGBench-V, and MADQA, DocPilot achieves strong gains with frontier VLM backbones, while DocPilot-4B delivers competitive performance against substantially larger models, demonstrating strong effectiveness and generalization across diverse long-document reasoning settings. The code is available at https://github.com/HKUSTDial/DocPilot.

cs.AI↗
arXiv2026

Can Agentic Trading Systems Pay for Their Own Intelligence?

Large language model (LLM) agents are increasingly used in trading systems, where model reasoning, tool use, and continual decisions incur costs that are expected to produce trading value. Existing evaluations typically report performance metrics, but rarely examine agentic viability: whether dynamic LLM-mediated decisions convert their induced costs into measurable incremental profit. To apply this criterion, we introduce TradeLens, a trace-grounded diagnostic toolkit for evaluating agentic trading systems from their trading records, runtime traces, and deployment configurations. It reconstructs trading trajectories, attributes profit and cost to interpretable evidence, and diagnoses whether and why an agent pays for its own intelligence. We conduct extensive analysis across backbone models, capital scales, trading frequencies, and system architectures, together with deployment discussion. Our results show that viability hinges on intelligence-to-profit conversion: models exhibit different failure patterns, such as poor asset selection in DeepSeek-V3.2 and negative timing in GLM-4.7, while capital scale, trading frequency, and architecture matter primarily in these runs, especially through their effects on decision-attributed timing value. These findings reframe the evaluation of LLMbased trading agents from capability-centric performance ranking to trace-grounded diagnosis of intelligence-to-profit conversion. Our code is available at https://github.com/ParadooxAI/TradeLens.

cs.AI↗
arXiv2026

Structured Composition of Verifiable Atomic Insights for Table-to-Report Generation

Table-to-report generation refers to the task of automatically generating article-level analyt- ical reports from relational tables and is an essential capability for automated data science and decision support. Its central challenge lies in systematically discovering verifiable com- posite insights across tables, attributes, and analytical perspectives, and organizing them into coherent, complete, and traceable evidence chains. Existing methods primarily rely on sequential, reactive data agents or direct Large Language Model(LLM) generation. They suffer from exploration bias: early local observations constrain subsequent actions, causing models to focus prematurely on local analyzes and miss cross-table or cross-dimensional evidence. We propose ComInsight, which reformulates insight discovery as the composition of atomic evidences. We first define an atomic insight as the smallest executable analytical unit conforming to a predefined analysis pattern and enumerate all valid atomic insights from database schema and content. These atoms are then organized into a multi-relational insight graph, where nodes represent verified data facts and edges encode logical, temporal, or hierarchical relations. Finally, a set of composition operators systematically fuses atomic nodes into higher-order composite conclusions. Every composite output is accompanied by executable SQL and fine-grained provenance, ensuring full verifiability. Across three benchmarks InsightBench, DDR-Bench, and T2R-Bench, ComInsight consistently outperforms strong baselines in factual correctness, novelty, and structural completeness. We believe ComInsight offers a reliable, efficient, and explainable path toward table-to-report generation.

cs.CL↗
arXiv2026

Raman Scattering Evidence for Fluctuating Kagome-Plane Moments in $\mathrm{B} \parallel [111]$ in Pr$_2$Zr$_2$O$_7$ pyrochlore

The exotic magnetic properties of the pyrochlore Pr$_2$Zr$_2$O$_7$ are determined by the coupling of spin, orbital, and lattice degrees of freedom. Magnetism in this material originates from the non-Kramers Pr$^{3+}$ ion, and the exotic behavior has been discussed both in the framework of dipole-quadrupole properties of magnetic moments and disorder splitting the non-Kramers doublet. We use magneto-Raman spectroscopy to probe the crystal electric field (CEF) excitations of Pr$^{3+}$ in magnetic fields up to 14 T applied along the [100] and [111] crystallographic directions at 2 K. For $\mathrm{B}\parallel[100]$, the field evolution of the Raman active crystal field modes is quantitatively described by conventional Zeeman splitting of the Pr$^{3+}$ ground state doublet, including the thermally populated upper Zeeman branch at elevated temperature. For $\mathrm{}{B}\parallel[111]$, the spectra separate into responses from the triangular and kagome sublattices, and the CEF excitations of Pr$^{3+}$ on the triangular lattice exhibit the expected Zeeman shift consistent with full moments polarization. In contrast, the CEF excitations of Pr$^{3+}$ on the kagome lattice do not split as expected and instead broaden strongly with increasing field. This behavior is inconsistent with a static polarized configuration and is captured phenomenologically by a motional-narrowing description in which the kagome plane moments fluctuate between Zeeman-split levels on a meV timescale. This result offers a natural explanation for the reduced magnetization observed in $\mathrm{}{B}\parallel[111]$ and provides further evidence of the dipole-quadrupole properties of magnetic moments. It highlights Raman scattering as a sensitive probe of exotic spin-orbital dynamics in frustrated magnets.

cond-mat.str-el↗
arXiv2026

Prioritized Rollouts for Efficient World Model-based Vision-Language-Action Policy Optimization

Vision-Language-Action (VLA) models have emerged as a powerful paradigm for embodied intelligence, but fine-tuning them with reinforcement learning (RL) remains constrained by the cost of real-world robot interaction. Model-based reinforcement learning (MBRL) reduces this cost by using a learned world model to generate rollouts for policy optimization. However, it becomes computationally expensive as VLA policies and world models scale. Existing methods typically treat states equally, overlooking substantial differences in their utility for policy improvement. In this paper, we show that policy uncertainty helps identify states with greater potential for policy improvement. The policy exhibits high uncertainty at only a small subset of states, often during decision-sensitive stages where small action differences can alter task outcomes, suggesting that policy improvements at these states could be particularly valuable. Building on these findings, we introduce U-GROW, a lightweight, plug-and-play sampling layer that directs more model rollouts to these informative states. By modifying only the branched-start distribution, U-GROW can be integrated into existing MBRL pipelines without changing the policy optimization objective. Experiments in both simulated and real-world manipulation tasks demonstrate the efficiency and effectiveness of U-GROW, supporting the use of policy uncertainty to guide experience generation.

cs.LG↗
arXiv2026

A Survey of Text-to-SQL in the Era of LLMs: Where are we, and where are we going?

Translating users' natural language queries (NL) into SQL queries (i.e., Text-to-SQL, a.k.a. NL2SQL) can significantly reduce barriers to accessing relational databases and support various commercial applications. The performance of Text-to-SQL has been greatly enhanced with the emergence of Large Language Models (LLMs). In this survey, we provide a comprehensive review of Text-to-SQL techniques powered by LLMs, covering its entire lifecycle from the following four aspects: (1) Model: Text-to-SQL translation techniques that tackle not only NL ambiguity and under-specification, but also properly map NL with database schema and instances; (2) Data: From the collection of training data, data synthesis due to training data scarcity, to Text-to-SQL benchmarks; (3) Evaluation: Evaluating Text-to-SQL methods from multiple angles using different metrics and granularities; and (4) Error Analysis: analyzing Text-to-SQL errors to find the root cause and guiding Text-to-SQL models to evolve. Moreover, we offer a rule of thumb for developing Text-to-SQL solutions. Finally, we discuss the research challenges and open problems of Text-to-SQL in the LLMs era. Text-to-SQL Handbook: https://github.com/HKUSTDial/NL2SQL_Handbook

cs.DB↗
arXiv2026

LakeHopper: Knowledge-Aware Adaptation of Column Type Annotators across Data Lakes

Column Type Annotation (CTA), which assigns a semantic type to a table column, underpins data integration, cleaning, and search over data lakes. State-of-the-art annotators are pre-trained language models (PLMs) fine-tuned on one particular corpus of tables, i.e., a source data lake, and they degrade sharply once deployed on a new (i.e., target) lake, whose tables and semantic type set both differ. Retraining per lake is prohibitive because it demands large volumes of expert annotations. We recast cross-lake adaptation as a knowledge management problem and make the resulting decomposition explicit: relative to a target annotator, a source annotator holds knowledge that must be discarded (source-specific), realigned and reused (shared), or acquired (target-specific). This decomposition exposes which part of the gap a general-purpose LLM can close and which part only target supervision can. Guided by it, we present LakeHopper, which adapts a source annotator under a fixed annotation budget through three coupled mechanisms: label-set realignment that transplants the output layer for shared types, LLM-verified gap discovery that localizes columns the annotator handles unreliably, and cluster-based propagation plus rehearsal fine-tuning that generalizes each flagged column into a labeling batch without erasing shared knowledge. Casting the LLM as a verifier of the annotator's own predictions rather than an annotator keeps every output inside the target type set, so LakeHopper structurally emits no out-of-domain labels, whereas prompted LLMs hallucinate types on 2.7-47.6% of columns. Across three data lake transfers of differing difficulty, LakeHopperlifts three PLM backbones by up to 71.4% relative macro-F1, reaches near-full-data quality with under 6% of the target labels, and matches fine-tuned table LLMs while training 27-131 times faster.

cs.CL↗
arXiv2026

EidosDoc: Implicit Structure Encoding for Cost-Effective Semi-Structured Document QA

Semi-structured documents are ubiquitous in scientific reports, financial statements, and technical manuals. Question answering over such documents requires simultaneous understanding of text, tables, charts, and complex hierarchical layouts. Existing methods either rely on repeatedly calling large language models for structure parsing and retrieval, leading to high cost and large latency, or they flatten the document and lose layout and hierarchy information, sacrificing answer accuracy. To address this, we propose EidosDoc, a novel system that achieves state-of-the-art accuracy with minimal computational expense. Our approach introduces three core innovations. (1) An Implicit Structure Encoder trained via contrastive learning and a structure consistency loss. This module jointly embeds hierarchical relationships, spatial positions, and textual content into a dense vector space, capturing document structure holistically without the need for manually defined and error-prone constructions. (2) A Hybrid Retrieval Pipeline that leverages BM25, layout fingerprints, and a lightweight cross-encoder to perform high-precision retrieval entirely without invoking an LLM, drastically reducing cost and latency. (3) A Dynamic Evidence Expansion mechanism that adaptively retrieves spatially adjacent and structurally related evidence, overcoming the evidence omission common in fixed-path retrieval methods. We evaluate EidosDoc on four benchmarks, and comprehensive evaluations show that EidosDoc achieves a new state-of-the-art accuracy on the four benchmarks. Crucially, it does so with a 50 times reduction in cost and 4 times lower latency compared to the previous state-of-the-art Method. These results demonstrate that EidosDoc establishes a new optimal trade-off among accuracy, cost, and speed, offering a practical and scalable path for accurate semi-structured document analysis.

cs.IR↗
arXiv2026

Beyond Tables: Doc2DB-Bench for Relationally Faithful Document-to-Database Construction

Practical AI systems increasingly need to turn long, heterogeneous documents into queryable relational databases, not isolated spreadsheets. In domains such as finance, healthcare, education, transportation, and enterprise operations, downstream workflows rely on normalized schemas, entity identities, keys, cross-table relationships, and integrity constraints for analytics, compliance, auditing, and SQL-backed decision making. Existing Document-to-Table benchmarks are insufficient for this setting: flattening evidence into single tables can duplicate entities, obscure many-to-many relationships, create sparse records, and avoid testing whether extracted facts form a valid database instance. This creates an urgent need to evaluate document understanding as database construction rather than field extraction. We introduce Doc2DB-Bench, a benchmark for Document-to-Database construction, containing 203 long-document instances across 42 schemas and seven domain groups, with 117 entity tables, 132 relationship tables, 7,341 rows, and 41,935 cells. Built through a controllable DB-to-Doc synthesis pipeline and organized by a taxonomy of intra-table extraction and inter-table reasoning, the generated documents undergo authenticity verification, proving indistinguishable from real-world references. Doc2DB-Bench thus provides a testbed for reliable, auditable, and relationally faithful LLM-based data systems. The benchmark is publicly available at https://github.com/SetonLiang/Doc2DB-Bench.

cs.CL↗
arXiv2026

Text2GraphQuery-Bench: A Text to Graph Query Benchmark

Graph models are fundamental to data analysis in domains rich with complex relationships. Unlike SQL, which benefits from a rel- atively unified standard and widespread familiarity, graph query languages are diverse (e.g., Cypher, GQL, SQL/PGQ) and far less fa- miliar to most users, making them significantly harder to learn and use. Text-to-Graph-Query systems address this barrier by trans- lating natural language into executable graph queries, enabling LLMs to serve as interfaces for Graph Database Management Systems (GDBMS). Existing benchmarks are limited in language coverage, rely on rigid synthesis, and lack comprehensive evaluation. We present Text2GraphQuery-Bench, the first benchmark covering all mainstream declarative property graph query languages (Cypher, GQL, and SQL/PGQ). It contains 267,276 (Question, Graph Query) pairs across 34 databases and 13 domains. Its construction supports adaptation from heterogeneous resources and domain-aware synthesis, while its Graph-IR-based design enables rapid extension to new languages. The evaluation protocol reports Grammar, GLEU, Similarity, and EX under graph-native difficulty, question abstraction, and schema aliasing. Experiments on 8 LLMs reveal: (i) a significant language gap exists - zero-shot GQL and SQL/PGQ Grammar is far below Cypher, yet few-shot prompting largely recovers it; (ii) fine-tuning an 8B model reaches or exceeds zero-shot large models, indicating unfamiliarity - rather than model capacity - is the primary barrier; (iii) as supervision increases, syntax errors recede, shifting bottlenecks to aggregation logic in GQL and schema linking in SQL/PGQ; (iv) higher question abstraction degrades EX due to intent-to-schema grounding issues, while schema aliasing has minimal impact; (v) EX consistently degrades from Easy to Extra Hard, with Extra Hard remaining a persistent bottleneck. *(Due to arXiv constraints, this abstract is shortened. See PDF for the full version.)*

cs.AI↗
arXiv2026

DataSpace: Benchmarking Data Agents for Verifiable Analytics over Heterogeneous Workspaces

Data agents enable natural-language analytics over organizational workspaces, where relevant evidence may be scattered across databases, structured files, long documents, and multimedia. Existing benchmarks largely isolate structured querying, retrieval, or open-ended analysis, leaving heterogeneous evidence discovery, complete tabular outputs, and deterministic evaluation insufficiently unified. We introduce DataSpace, a benchmark in which data agents produce verifiable tabular results from task-local heterogeneous workspaces. It contains 410 cross-language tasks and 7,439 artifacts totaling 15.01 GB across CSV, JSON, SQLite, Markdown, PDF, and video. DataSpace also served as the official evaluation benchmark for the KDD Cup 2026 Data Agents for Complex Data Analysis competition. Each agent receives only a question and workspace and returns the complete requested tabular result. We construct DataSpace with DataSpace-Builder, an execution-grounded framework comprising cross-language transformation, constraint-aware relational sampling, modality routing and artifact rendering, and human review and task repair by 11 domain experts. A deterministic evaluator performs header-invariant column alignment, type- and precision-aware normalization, and order-aware row comparison. Across six recently released frontier multimodal models and five widely used agent harnesses, the best accuracy reaches 66.34%, while harness choice creates a 15.36-point spread with the backbone fixed. Multimodal evidence integration and joins consistently reduce accuracy across all six backbones. These results show that DataSpace remains unsaturated and identify key challenges for improving data-agent reliability.

cs.AI↗
arXiv2026

Monte Carlo Tree Search for Table-to-Multimodal Report Generation

Automatically generating professional multimodal reports comprising both textual analysis and visual charts from structured tabular data is a critical challenge in data intelligence. Existing methods suffer from fixed linear pipelines and isolated subtask processing, which hinder joint optimization of factual accuracy, visual quality, and narrative coherence. To address these issues, this paper proposes MCTS-Report, a Monte Carlo Tree Search (MCTS)-driven framework that formulates multimodal table-to-report generation as a progressive construction process over a structured search space. The core idea is to decompose report generation into atomic actions, including chapter planning, visualization task identification, chart generation, insight organization, and narrative refinement, each executed by an LLM based on dynamic reasoning conditioned on the current report state. We use an LLM to generate step-by-step reasoning and actions during MCTS, storing the reasoning trajectory in each node for context-aware, coherent report construction. To guide the search, we design a multi-dimensional reward function that jointly evaluates numerical fact consistency (via SQL), chart quality, chart-text alignment, and structural completeness, while incorporating a diversity penalty to suppress repeated charts and a precondition check to prune invalid actions. We also construct MMRBench, a comprehensive benchmark comprising real-world tables from six domains, paired with expert-refined reference report structures and verifiable key insights. Experiments on MMRBench demonstrate that MCTS-Report significantly outperforms strong baselines across structural completeness, numerical accuracy, chart-text alignment, and insight novelty, achieving a 77.9 overall score.

cs.AI↗
arXiv2026

InteractComp: Evaluating Search Agents With Ambiguous Queries

Language agents have demonstrated remarkable potential in web search and information retrieval. However, many search-agent benchmarks assume that user queries are complete and unambiguous. This assumption leaves under-tested a practical failure mode: agents may face ambiguous requests where the intended target cannot be identified without clarification. Yet most agents lack interactive mechanisms during the search process, and existing benchmarks cannot assess this capability. To address this gap, we introduce InteractComp, a benchmark designed to evaluate whether search agents can recognize query ambiguity and actively interact to resolve it during search. Following the principle of easy to verify, interact to disambiguate, we construct 210 expert-curated questions across 9 domains through a target-distractor methodology that creates controlled ambiguity resolvable only through interaction. Evaluation of 17 models reveals striking failure: the best model achieves only 13.73% accuracy despite 71.50% with complete context, exposing systematic overconfidence rather than reasoning deficits. Forced interaction produces dramatic gains, demonstrating latent capability current strategies fail to engage. Longitudinal analysis shows interaction capabilities stagnated over 15 months while search performance improved seven-fold, revealing a critical blind spot. This stagnation, coupled with the immediate feedback inherent to search tasks, makes InteractComp a valuable resource for both evaluating and training interaction capabilities in search agents. The code is available at https://github.com/FoundationAgents/InteractComp.

cs.CL↗
arXiv2026

AnnoRetrieve: Efficient Structured Retrieval for Unstructured Document Analysis

Unstructured documents dominate enterprise and web data, but their lack of explicit organization hinders precise information retrieval. Current mainstream retrieval methods, especially embedding-based vector search, rely on coarse-grained semantic similarity, incurring high computational cost and frequent LLM calls for post-processing. To address this critical issue, we propose AnnoRetrieve, a novel retrieval paradigm that shifts from embeddings to structured annotations, enabling precise, annotation-driven semantic retrieval. Our system replaces expensive vector comparisons with lightweight structured queries over automatically induced schemas, dramatically reducing LLM usage and overall cost. The system integrates two synergistic core innovations: SchemaBoot, which automatically generates document annotation schemas via multi-granularity pattern discovery and constraint-based optimization, laying a foundation for annotation-driven retrieval and eliminating manual schema design, and Structured Semantic Retrieval (SSR), the core retrieval engine, which unifies semantic understanding with structured query execution; by leveraging the annotated structure instead of vector embeddings, SSR achieves precise semantic matching, seamlessly completing attribute-value extraction, table generation, and progressive SQL-based reasoning without relying on LLM interventions. This annotation-driven paradigm overcomes the limitations of traditional vector-based methods with coarse-grained matching and heavy LLM dependency and graph-based methods with high computational overhead. Experiments on three real-world datasets confirm that AnnoRetrieve significantly lowers LLM call frequency and retrieval cost while maintaining high accuracy. AnnoRetrieve establishes a new paradigm for cost-effective, precise, and scalable document analysis through intelligent structuring.

cs.IR↗
arXiv2026

CSTrader: A Testbed for Language-Grounded Trading in a Community-Driven Virtual Asset Market

Niche asset markets, such as Counter-Strike 2 (CS2) weapon skins, are small, volatile, and heavily driven by community discussions and platform rules. These properties make them hard for traditional quantitative models, but provide an ideal testbed for studying how large language models (LLMs) turn unstructured text into trading actions. We present CSTrader, a multi-agent framework for language-grounded trading in the CS2 skin market. The system first integrates heterogeneous signals from various sources, then uses specialized agents for technical analysis, liquidity, events, and (reversed) sentiment, and finally applies risk control, transaction friction, and portfolio management agents to produce buy, sell, or hold decisions under realistic trading frictions. We build a live-like evaluation environment with real CS2 data from a highly volatile period and evaluate several recent LLM backbones. Across models, CSTrader consistently outperforms both a falling market index (-15.62%) and simple single-prompt LLM baselines, achieving up to a 7.58% cumulative return with controlled risk. Ablation studies show that liquidity, reversed sentiment, and transaction friction agents are crucial for turning noisy language signals into stable profits, suggesting that niche, language-driven markets are a useful benchmark for future language-to-action research. Code is available at: https://github.com/IatomicreactorI/CSGOTrading?tab=readme-ov-file#quick-start

cs.AI↗
arXiv2026

A Risk Decomposition Framework for Pre-Hoc Fine-Tuning Prediction

The high cost of fine-tuning LLMs poses a significant economic barrier; pre-hoc performance prediction offers a critical solution to substantially reduce this expense. However, the theoretical limits of pre-hoc performance prediction remain unexplored. We formulate it as a stochastic estimation problem under information constraints, decomposing prediction risk into two components: an intrinsic limit (static data-model compatibility) and a reducible optimization variance. We prove that optimization variance admits a necessary lower bound on its decay rate, implying fundamental constraints on how quickly uncertainty dissipates, regardless of the predictor used. Based on these dynamics, we derive a budget-optimal probing principle and introduce a predictability phase diagram that organizes tasks into three distinct regimes: Static-Sufficient, Dynamic-Critical, and Noise-Dominant. Extensive experiments on synthetic and real-world benchmarks validate these theoretical regimes and demonstrate the efficiency of our probing strategy.

cs.LG↗