How Code Representation Shapes False-Positive Dynamics in Cross-Language LLM Vulnerability Detection

• URL: http://arxiv.org/abs/2604.27714v1
• Date: Thu, 30 Apr 2026 11:01:03 GMT
• Title: How Code Representation Shapes False-Positive Dynamics in Cross-Language LLM Vulnerability Detection
• Abstract summary: How code representation format shapes false positive behaviour in cross-language vulnerability detection remains poorly understood.<n>We systematically vary training intensity and code representation format, comparing raw source text with pruned Abstract Syntax Trees at both training time and inference time.<n>Cross-language FPR reflects the joint effect of training-time and inference-time representation, not either alone.
• Score: 8.360057179027322
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: How code representation format shapes false positive behaviour in cross-language LLM vulnerability detection remains poorly understood. We systematically vary training intensity and code representation format, comparing raw source text with pruned Abstract Syntax Trees at both training time and inference time, across two 8B-parameter LLMs (Qwen3-8B and Llama 3.1-8B-Instruct) fine-tuned on C/C++ data from the NIST Juliet Test Suite (v1.3) and evaluated on Java (OWASP Benchmark v1.2) and Python (BenchmarkPython v0.1). Cross-language FPR reflects the joint effect of training-time and inference-time representation, not either alone. Text fine-tuning drives FPR upward monotonically (Qwen3-8B: 0.763 zero-shot, 0.866 pilot, 1.000 full-scale) while F1 remains stable (0.637-0.688), masking the collapse. We argue surface-cue memorisation is the primary mechanism: text fine-tuning encodes C/C++-specific API names and syntactic idioms as vulnerability triggers that fire indiscriminately on target-language code. A cross-representation probe, applying text-trained weights to AST-encoded input without retraining, isolates this: Qwen3-8B FPR drops from 0.866 to 0.583, and 37.2% of false positives revert to true negatives under AST input alone. Direct AST fine-tuning does not preserve the benefit (FPR at least 0.970), as flat linearisation introduces structural surface cues of its own. The pattern replicates across both model families. On BenchmarkPython the AST probe yields FPR=0.554, within 2.9 percentage points of the Java result, despite maximal surface-syntax differences, substantially weakening a domain-shift explanation. These findings motivate a pre-deployment consistency gate, running alerts through both text and AST paths, as a retraining-free filter for false-positive-sensitive settings, at the cost of reduced recall.

Related papers

• PolicyGuard: Prompt-Configurable Semantic DLP for LLM Coding Agents [0.8836202788658859]
  Existing data loss prevention solutions rely on rigid patterns, model fine-tuning, or vendor-managed classifiers with customizability.<n>We present PolicyGuard, a pre-model interception framework that classifies prompts using a natural language policy file.<n>Our key contributions are: (1) the policy-as-prompt paradigm, where classification criteria are defined entirely in a policy document editable by non-engineers without code changes or model retraining; and (2) a sealed evaluation protocol with template-family-level data splits, hidden holdouts, and frozen test sets to rigorously assess generalization.
  arXiv  Detail & Related papers  (2026-08-03T08:10:15Z)
• PCAP-LM: An LLM-Native Text Representation for TLS Bulk Traffic Analysis [0.0]
  We present PCAP-LM, a flow-centric, LLM-native text representation that acts as a lossy knowledge extraction step.<n>Raw captures are transcoded into semantic summaries using PacketGlyphs.<n>The BPE vocabulary fully saturates at 159 tokens, achieving an 812x size reduction over tshark -V.
  arXiv  Detail & Related papers  (2026-07-30T12:10:12Z)
• Weave of Formal Thought [51.56484100374058]
  We introduce Weave of Formal Thought (WoFT), a paradigm uniting rigorous syntactic validation with learned structural representations.<n>We present a latent-variable fine-tuning method training the language model to interleave non-terminal grammar symbols directly into generation.<n>For Python, fine-tuning StarCoder2-3B with our RWS objective reduces per-token cross-entropy by 14.3% relative to a text-only SFT baseline.
  arXiv  Detail & Related papers  (2026-06-24T15:58:11Z)
• Context-Aware Distillation and Ablation for Text2DSL [78.29352305480285]
  We replace prompt-only synthetic generation with context-aware distillation.<n>A teacher large language model operates under an explicitly defined structured context.<n>This scales the verified PolkitBench corpus from 4,204 to 10,073 pairs.
  arXiv  Detail & Related papers  (2026-06-21T16:27:24Z)
• Closing the Modality Gap in Zero-Shot HAR: Contrastive Training and Separability-Optimized Prototypes on IMU Data [0.0]
  We evaluate seven configurations combining three inference methods with two training pipelines on the PAMAP2 dataset.<n>We find that the modality gap is a training-time phenomenon governed by the objective.<n>We also demonstrate that overall accuracy is a misleading primary metric when test-set class distributions are imbalanced.
  arXiv  Detail & Related papers  (2026-06-09T12:39:41Z)
• A Protocol-Language Model for Network Intrusion (Without Deep Packet Inspection) [0.0]
  We present PLM-NIDS, which proves three claims in sequence.<n>Because it reads only IP/ TCP/UDP headers, it is inherently encryption-agnostic.
  arXiv  Detail & Related papers  (2026-05-29T07:03:11Z)
• Recursive language models for jailbreak detection: a procedural defense for tool-augmented agents [0.0]
  We present RLM-JB, an end-to-end jailbreak detection framework built on Recursive Language Models (RLMs)<n>RLM-JB treats detection as a procedure rather than a one-shot classification.<n>On AutoDAN-style adversarial inputs, RLM-JB achieves high detection effectiveness across three LLM backends.
  arXiv  Detail & Related papers  (2026-02-18T15:07:09Z)
• LLM-Driven Adaptive Source-Sink Identification and False Positive Mitigation for Static Analysis [0.0]
  textscAdaTaint adaptively infers source/sink specifications and filters spurious alerts through neuro-symbolic reasoning.<n>textscAdaTaint grounds model suggestions in program facts and constraint validation, ensuring both adaptability and determinism.<n>Results show that textscAdaTaint reduces false positives by textbf43.7% on average and improves recall by textbf11.2% compared to state-of-the-art baselines.
  arXiv  Detail & Related papers  (2025-11-06T03:44:10Z)
• PrefixNLI: Detecting Factual Inconsistencies as Soon as They Arise [60.63315470285562]
  MiniTruePrefixes is a novel specialized model that better detects factual inconsistencies over text prefixes.<n>We show that integrating MiniTruePrefixes into a controlled decoding framework substantially improves factual consistency in abstractive summarization.
  arXiv  Detail & Related papers  (2025-11-03T09:07:44Z)
• From Flows to Words: Can Zero-/Few-Shot LLMs Detect Network Intrusions? A Grammar-Constrained, Calibrated Evaluation on UNSW-NB15 [0.41998444721319217]
  Large Language Models (LLMs) can reason over natural-language inputs, but their role in intrusion detection without fine-tuning remains uncertain.<n>This study evaluates a promptonly approach by converting each network flow to a compact textual record and augmenting it with lightweight, domain-inspired flags.<n>We compare zero-shot, instruction-guided, and fewshot prompting to strong neural baselines under identical splits, reporting accuracy, precision, recall, F1, and macro scores.
  arXiv  Detail & Related papers  (2025-10-18T02:11:50Z)
• RepreGuard: Detecting LLM-Generated Text by Revealing Hidden Representation Patterns [50.401907401444404]
  Large language models (LLMs) are crucial for preventing misuse and building trustworthy AI systems.<n>We propose RepreGuard, an efficient statistics-based detection method.<n> Experimental results show that RepreGuard outperforms all baselines with average 94.92% AUROC on both in-distribution (ID) and OOD scenarios.
  arXiv  Detail & Related papers  (2025-08-18T17:59:15Z)
• Who Wrote This? The Key to Zero-Shot LLM-Generated Text Detection Is GECScore [51.65730053591696]
  We propose a simple yet effective black-box zero-shot detection approach based on the observation that human-written texts typically contain more grammatical errors than LLM-generated texts.<n> Experimental results show that our method outperforms current state-of-the-art (SOTA) zero-shot and supervised methods.
  arXiv  Detail & Related papers  (2024-05-07T12:57:01Z)
• ASPIRO: Any-shot Structured Parsing-error-Induced ReprOmpting for Consistent Data-to-Text Generation [0.0]
  ASPIRO is an approach for structured data verbalisation into short template sentences in zero to few-shot settings. Unlike previous methods, our approach prompts large language models to directly produce entity-agnostic templates.
  arXiv  Detail & Related papers  (2023-10-27T03:39:51Z)
• LeTI: Learning to Generate from Textual Interactions [60.425769582343506]
  We explore LMs' potential to learn from textual interactions (LETI) that not only check their correctness with binary labels but also pinpoint and explain errors in their outputs through textual feedback. Our focus is the code generation task, where the model produces code based on natural language instructions. LETI iteratively fine-tunes the model, using the objective LM, on a concatenation of natural language instructions, LM-generated programs, and textual feedback.
  arXiv  Detail & Related papers  (2023-05-17T15:53:31Z)
• DetectGPT: Zero-Shot Machine-Generated Text Detection using Probability Curvature [143.5381108333212]
  We show that text sampled from an large language model tends to occupy negative curvature regions of the model's log probability function. We then define a new curvature-based criterion for judging if a passage is generated from a given LLM. We find DetectGPT is more discriminative than existing zero-shot methods for model sample detection.
  arXiv  Detail & Related papers  (2023-01-26T18:44:06Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.