How Close is ChatGPT to Human Experts? Comparison Corpus, Evaluation, and Detection

cs.CL 2026-05 unverdicted novelty 7.0 of 10

A weighted change point detection algorithm locates human- and LLM-written segments in mixed text and attains minimax-optimal localization under heterogeneous sentence scores.

cs.CL 2026-07 reject novelty 6.0 of 10

A training-free detector using discrete wavelet analysis of token log-probability sequences reports AUROC 0.99/0.85/0.75 on HC3/M4/MAGE, but configuration selection on the test split weakens the numbers.

cs.CL 2026-07 conditional novelty 6.0 of 10

LLMs overuse the 'not X, but Y' self-correction pattern in persuasive registers and underuse it in informal Q&A; a prompt or a detachable LoRA dial adjusts it to human levels.

cs.CL 2026-01 conditional novelty 6.0 of 10

Adding human-alignment augmentation (roleplaying, BPO, self-refine, RLDF) to machine-generated text both fools existing detectors and improves the generalization of detectors fine-tuned on it.

cs.LG 2025-10 unverdicted novelty 6.0 of 10

SurpMark detects machine-generated text by estimating state-transition matrices from discretized surprisals and scoring them with generalized Jensen-Shannon divergence to human versus machine references.

cs.CY 2025-09 conditional novelty 6.0 of 10

After ChatGPT's launch, Stack Overflow posts became longer and shifted toward medium-difficulty questions, suggesting users turn to the crowd for more complex problems.

cs.CL 2025-08 conditional novelty 6.0 of 10

CoCoNUTS is a six-mode peer-review benchmark and CoCoDet a multi-task detector that classifies reviews by content origin, reaching 98% macro F1 in-domain.

cs.CL 2025-08 conditional novelty 6.0 of 10

An explanation-then-detection framework for machine-generated text in dialogues that uses dialogue acts and token attributions to produce fast, non-expert-friendly explanations.

quant-ph 2025-08 unverdicted novelty 6.0 of 10

Relativistic motion of the reservoirs in a three-level maser is claimed to yield a generalized Carnot bound that allows efficiency above the ordinary Carnot limit.

cs.CL 2025-08 conditional novelty 6.0 of 10

A detector that projects a text's hidden neural activations onto a direction learned from human versus AI writing differences reports higher accuracy than baselines, including on unseen generators and attacked texts.

cs.CL 2025-07 conditional novelty 6.0 of 10

SHIELD shows that standard AUROC overstates AI-text detector quality, and that six zero-shot detectors collapse under a controllable word-replacement humanification.

cs.SE 2025-07 conditional novelty 6.0 of 10

CodeGPTSensor+, trained with adversarial samples that combine identifier renaming and structure transformation, is substantially more robust to subtle modifications of LLM-generated code than the original CodeGPTSensor.

cs.CL 2025-06 conditional novelty 6.0 of 10

PhantomHunter detects text from privately fine-tuned LLMs by learning shared token-probability traits within LLaMA, Gemma and Mistral families, reporting F1 above 96% on held-out derivatives.

cs.CL 2025-06 conditional novelty 6.0 of 10

A new LLM-text detector built from twin memory networks and domain-generalization losses outperforms prior detectors on held-out LLMs and domains in the paper's benchmark.

cs.CL 2025-06 conditional novelty 6.0 of 10

Fine-tuned LLMs generate social media text that evades state-of-the-art detectors and human readers, dropping detection accuracy from up to 99.9% to near chance.

cs.CL 2025-06 conditional novelty 6.0 of 10

DivScore detects AI-written medical and legal text by dividing a domain-tuned model's entropy by its disagreement with a general model, beating baselines on a new benchmark.

cs.CL 2025-06 conditional novelty 6.0 of 10

Human involvement in AI-generated academic text can be estimated continuously by training a RoBERTa regressor on BERTScore-derived labels, outperforming binary detectors on a new synthetic dataset.

cs.CL 2025-06 conditional novelty 6.0 of 10

Current machine-generated text detectors, especially metric-based ones, perform poorly on word-level detection in coauthored texts, while finetuned DeBERTa achieves strong but imperfect performance.

cs.CL 2025-05 conditional novelty 6.0 of 10

Arabic text written by LLMs carries detectable stylometric signatures, and fine-tuned XLM-RoBERTa detectors reach near-perfect F1 on academic abstracts but degrade on social media.

cs.CR 2025-04 conditional novelty 6.0 of 10

Q-faker generates adversarial examples for NLU classifiers using surrogate-model gradients and controlled generation, requiring zero queries to the target.

cs.LG 2025-02 conditional novelty 6.0 of 10

A three-method auditing framework detects with roughly 87 to 97 percent accuracy whether classifiers, generators, and t-SNE plots were trained on or derived from LLM-generated synthetic data.

cs.CL 2025-01 conditional novelty 6.0 of 10

A new 171-speech Greek campaign corpus with 31,674 human-validated paragraph annotations for six NLP tasks, plus a ChatGPT-versus-expert accuracy analysis.

cs.CL 2025-01 conditional novelty 6.0 of 10

A learned token-weighting head over next-token probabilities improves AI-text detection accuracy and generalization with about one million trainable parameters.

cs.AI 2024-12 conditional novelty 6.0 of 10

A detector trained on a new multi-LLM social media benchmark estimates that AI-generated text on Medium and Quora grew from about 2% to roughly 37 to 39 percent between 2022 and 2024, while Reddit stayed near 2%.

cs.AI 2024-12 conditional novelty 6.0 of 10

A new 749K-sample academic-writing benchmark shows that supervised detectors outperform metric-based ones, attribution is difficult for metric methods, and adapting to new LLMs leaves a sizable gap.

cs.CL 2024-11 conditional novelty 6.0 of 10

ChatGPT-3.5-generated State of the Union speeches are stylistically distinguishable from real presidential addresses across word, sentence, part-of-speech, and rhetorical measures.

cs.LG 2026-07 conditional novelty 5.0 of 10

ASK-NN, a one-sided nearest-neighbor coincidence test, detects distribution drift between reference and query samples with theoretical guarantees and competitive empirical performance on LLM hallucination and artifici...

cs.CL 2026-07 conditional novelty 5.0 of 10

A sliding-window, trajectory-difference contrastive learner beats six AI-text detectors on RAID, NYT-AI, and OpenReview reviews.

cs.CL 2025-08 conditional novelty 5.0 of 10

LLM-generated jailbreak prompts elicited health misinformation from GPT-3.5, Llama 3.1-8B, and Gemini 2.0 Flash at high rates, and both LLM judges and simple classifiers detected the resulting texts with high accuracy.

cs.CL 2025-07 conditional novelty 5.0 of 10

Stylometric features and tree-based classifiers separate human-written Wikipedia summaries from LLM-generated texts with high cross-validated accuracy on a new seven-class benchmark, though performance drops on other ...

cs.CL 2025-05 conditional novelty 5.0 of 10

ORPO fine-tuning, which explicitly penalizes disfavored language-mixed responses, nearly eliminates language confusion in Korean-generation LLMs without hurting QA accuracy.

cs.CL 2025-05 conditional novelty 5.0 of 10

A conformal-prediction wrapper with length-binned thresholds bounds the false-positive rate of machine-generated-text detectors at a user-chosen level alpha while improving true-positive rate in experiments.

cs.AI 2025-04 conditional novelty 5.0 of 10

LLMs given numerical arousal and valence coordinates produce text that sentiment analysis places in the same region of Russell's circumplex, demonstrating limited but real control over emotional tone.

cs.IR 2025-02 conditional novelty 5.0 of 10

A three-stage RAG pipeline generates a cited summary (GenText) from PubMed Central passages and ranks health documents by topical relevance plus alignment with that summary, outperforming baselines on CLEF eHealth and...

cs.CL 2025-01 conditional novelty 5.0 of 10

A COLING 2025 shared task benchmark showing that current machine-generated text detectors reach only moderate accuracy and degrade badly on out-of-domain and humanized AI text.

cs.IR 2024-11 conditional novelty 5.0 of 10

LLM-generated Boolean queries for systematic reviews are unstable across seeds, and the original ChatGPT results could not be reproduced with the documented setup.

cs.SE 2024-11 conditional novelty 5.0 of 10

Pre-trained models fine-tuned on LLM-generated data are less robust to adversarial attacks than models fine-tuned on human-written data in three software analytics tasks.

cs.CL 2026-08 conditional novelty 4.0 of 10

EchoPrompt detects LLM-generated text by measuring the likelihood gain from restoring a generic assistant-style prompt, calibrated against a base model.

stat.ML 2026-07 conditional novelty 4.0 of 10

The paper develops joint casewise and cellwise MCD estimators for multi-class robust covariance estimation and applies them within a Mahalanobis-distance-based detection framework across four AI-content detection task...

cs.CL 2025-10 reject novelty 4.0 of 10

A dataset of ~58k NYT articles plus AI rewrites from six LLMs, evaluated with a rewrite-distance baseline reaching 58.35% detection and 8.92% attribution accuracy.

cs.CL 2025-09 conditional novelty 4.0 of 10

MoSEs improves AI-text detection by replacing a single score threshold with a threshold learned from style-matched reference texts and linguistic statistics.

cs.CY 2025-06 reject novelty 4.0 of 10

LLM replies to climate change posts are more emotionally neutral and lower in intensity than the human posts they respond to.

cs.CY 2025-05 conditional novelty 4.0 of 10

A narrative review that aligns generative AI tools with formative assessment principles, provides classroom prompt examples, and identifies missing evaluation metrics for AI feedback.

cs.IR 2025-05 conditional novelty 4.0 of 10

A reranker fine-tuned on hard negatives selected by two cosine-distance criteria outperforms older negative sampling methods on enterprise and domain-specific retrieval benchmarks.

cs.CL 2025-05 reject novelty 4.0 of 10

LM2OTIFS uses word co-occurrence graphs and GNNExplainer to detect and explain machine-generated text, with strong in-domain accuracy but unsupported faithfulness claims and a flawed theoretical proof.

cs.CL 2025-04 conditional novelty 4.0 of 10

DP-Net uses a DDPG reinforcement-learning agent to dynamically tune Gaussian perturbation of embeddings, improving average cross-domain accuracy and adversarial robustness of AI-generated text detection.

cs.CL 2024-12 conditional novelty 4.0 of 10

A bilingual shared-task benchmark shows that fine-tuned classifiers distinguish AI-generated from human-written academic essays with macro-F1 above 0.98 within the challenge's test sets.

cs.CL 2024-12 conditional novelty 4.0 of 10

GPT-generated State of the Union speeches differ stylistically from real presidential speeches, with more "we", more positive and abstract terms, and lower authenticity.

cs.LG 2024-12 conditional novelty 4.0 of 10

A reinforcement-learning router that converges to one LLM per query session, reducing cost and latency while keeping answer quality competitive.

cs.CL 2024-12 conditional novelty 4.0 of 10

Human-written texts show higher linguistic variability, richer vocabulary and emotional content, and lower syntactic depth than texts from five LLMs.

cs.CL 2025-07 conditional novelty 3.0 of 10

A non-neural stylometric detector using LightGBM on spaCy-derived features reached a final mean score of 0.897 on the PAN 2025 task, below the 0.922 TF-IDF SVM baseline.

cs.CV 2025-02 unverdicted novelty 3.0 of 10

A survey organizing AI-generated media detection into Non-MLLM and MLLM based methods, with task and benchmark taxonomies.

cs.CL 2025-01 unverdicted novelty 3.0 of 10

A broad survey of LLM methods and applications, plus an empirical section on how code-rich pretraining may influence chain-of-thought reasoning, the details of which are not visible in the supplied text.