On the Limits of LLM Adaptability: Impact of Model-Internalized Priors on Annotation Task Performance

• URL: http://arxiv.org/abs/2606.00467v1
• Date: Sat, 30 May 2026 01:21:14 GMT
• Title: On the Limits of LLM Adaptability: Impact of Model-Internalized Priors on Annotation Task Performance
• Abstract summary: We investigate how model-internalized priors interact with user-provided instructions.<n>We find that nearly two-thirds of zero-shot errors are resistant to correction.<n>High-confidence errors prove especially resistant to correction.
• Score: 0.7303825588015752
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large Language Models (LLMs) are increasingly used for zero-shot annotation and LLM-as-a-judge tasks, yet their reliability hinges on how model-internalized priors interact with user-provided instructions. We investigate three dimensions of this interaction: (1) how an LLM's familiarity with data and task definitions affects performance, (2) the extent to which additional information in prompts can correct zero-shot errors ("decision stickiness"), and (3) model susceptibility to misaligned task definitions. Through experiments on toxicity detection across diverse datasets (spanning social media, gaming, news, and forums) using both dense and mixture-of-experts models, we find that nearly two-thirds of zero-shot errors are resistant to correction, with an overall rescue rate (fraction of initial errors corrected by prompting) of only 34.8%. High-confidence errors prove especially resistant to correction. When given misaligned definitions, LLMs follow them while maintaining confidence levels unchanged from the aligned condition. Crucially, we introduce Definition-Specific Familiarity (DSF), which measures alignment between a model's internal concept and the task definition. After controlling for dataset-level confounds, DSF shows a positive association with model performance (partial r = +0.41), while three distinct memorization metrics (ROUGE-L, BERTScore, and embedding cosine similarity) all fail to show a positive association. These findings show the limitations of prompt-based correction in annotation tasks, highlighting the importance of definition alignment over text-level memorization.

Related papers

• From Passive Response to Proactive Correction: Enhancing LLM Robustness Against Input Fact Perturbations [12.021107313829082]
  Large language models (LLMs) produce confident yet factually incorrect responses when user inputs contain misleading premises.<n>We propose DEDUCE, a three-stage framework that transforms LLMs from passive responders into proactive error correctors.<n> Experiments on TruthfulQA, FalseQA, and our MisFactQA benchmark demonstrate that DEDUCE significantly improves both accuracy and error correction capability.
  arXiv  Detail & Related papers  (2026-08-26T15:13:50Z)
• PluginEval: A Diagnostic Benchmark for Fine-Grained Error Attribution in Function Calling [11.1017602726296]
  Large Language Models increasingly operate as autonomous agents.<n>Current benchmarks face three structural limitations.<n>Eval is a benchmark constructed through a two-stage framework.
  arXiv  Detail & Related papers  (2026-08-09T13:25:54Z)
• 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)
• LLM Microscope: What Model Internals Reveal About Answer Correctness and Context Utilization [9.410181019585822]
  We operationalize interpretability methods to ascertain whether we can predict the correctness of model outputs.<n>We consider correct, incorrect, and irrelevant context and introduce metrics to distinguish amongst them.<n>Our model-internals-based metric significantly outperforms prompting baselines at distinguishing between correct and incorrect context.
  arXiv  Detail & Related papers  (2025-10-05T03:14:05Z)
• Everything You Wanted to Know About LLM-based Vulnerability Detection But Were Afraid to Ask [30.819697001992154]
  Large Language Models are a promising tool for automated vulnerability detection.<n>Despite widespread adoption, a critical question remains: Are LLMs truly effective at detecting real-world vulnerabilities?<n>This paper challenges three widely held community beliefs: that LLMs are (i) unreliable, (ii) insensitive to code patches, and (iii) performance-plateaued across model scales.
  arXiv  Detail & Related papers  (2025-04-18T05:32:47Z)
• SPARC: Score Prompting and Adaptive Fusion for Zero-Shot Multi-Label Recognition in Vision-Language Models [74.40683913645731]
  Zero-shot multi-label recognition (MLR) with Vision-Language Models (VLMs) faces significant challenges without training data, model tuning, or architectural modifications.<n>Our work proposes a novel solution treating VLMs as black boxes, leveraging scores without training data or ground truth.<n>Analysis of these prompt scores reveals VLM biases and AND''/OR' signal ambiguities, notably that maximum scores are surprisingly suboptimal compared to second-highest scores.
  arXiv  Detail & Related papers  (2025-02-24T07:15:05Z)
• Label Convergence: Defining an Upper Performance Bound in Object Recognition through Contradictory Annotations [0.0]
  We introduce the notion of "label convergence" to describe the highest achievable performance under the constraint of contradictory test annotations.<n>We approximate that label convergence is between 62.63-67.52 mAP@[0.5:0.95:0.05] for LVIS with 95% confidence, attributing these bounds to the presence of real annotation errors.<n>With current state-of-the-art (SOTA) models at the upper end of the label convergence interval for the well-studied LVIS dataset, we conclude that model capacity is sufficient to solve current object detection problems.
  arXiv  Detail & Related papers  (2024-09-14T10:59:25Z)
• Evaluating Generative Language Models in Information Extraction as Subjective Question Correction [49.729908337372436]
  We propose a new evaluation method, SQC-Score. Inspired by the principles in subjective question correction, we propose a new evaluation method, SQC-Score. Results on three information extraction tasks show that SQC-Score is more preferred by human annotators than the baseline metrics.
  arXiv  Detail & Related papers  (2024-04-04T15:36:53Z)
• Attention Satisfies: A Constraint-Satisfaction Lens on Factual Errors of Language Models [38.79074982172423]
  We investigate the internal behavior of Transformer-based Large Language Models (LLMs) when they generate factually incorrect text. We propose modeling factual queries as constraint satisfaction problems. We find a strong positive relationship between the LLM's attention to constraint tokens and the factual accuracy of generations.
  arXiv  Detail & Related papers  (2023-09-26T17:48:55Z)
• Improving Open Information Extraction with Large Language Models: A Study on Demonstration Uncertainty [52.72790059506241]
  Open Information Extraction (OIE) task aims at extracting structured facts from unstructured text. Despite the potential of large language models (LLMs) like ChatGPT as a general task solver, they lag behind state-of-the-art (supervised) methods in OIE tasks.
  arXiv  Detail & Related papers  (2023-09-07T01:35:24Z)
• Zero-shot Faithful Factual Error Correction [53.121642212060536]
  Faithfully correcting factual errors is critical for maintaining the integrity of textual knowledge bases and preventing hallucinations in sequence-to-sequence models. We present a zero-shot framework that formulates questions about input claims, looks for correct answers in the given evidence, and assesses the faithfulness of each correction based on its consistency with the evidence.
  arXiv  Detail & Related papers  (2023-05-13T18:55:20Z)
• BUMP: A Benchmark of Unfaithful Minimal Pairs for Meta-Evaluation of Faithfulness Metrics [70.52570641514146]
  We present a benchmark of unfaithful minimal pairs (BUMP) BUMP is a dataset of 889 human-written, minimally different summary pairs. Unlike non-pair-based datasets, BUMP can be used to measure the consistency of metrics.
  arXiv  Detail & Related papers  (2022-12-20T02:17:30Z)

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.