Spotting LLMs With Binoculars: Zero-Shot Detection of Machine-Generated Text
Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel pith:7SNFBXEIrecord.jsonopen to challenge →
read the original abstract
Detecting text generated by modern large language models is thought to be hard, as both LLMs and humans can exhibit a wide range of complex behaviors. However, we find that a score based on contrasting two closely related language models is highly accurate at separating human-generated and machine-generated text. Based on this mechanism, we propose a novel LLM detector that only requires simple calculations using a pair of pre-trained LLMs. The method, called Binoculars, achieves state-of-the-art accuracy without any training data. It is capable of spotting machine text from a range of modern LLMs without any model-specific modifications. We comprehensively evaluate Binoculars on a number of text sources and in varied situations. Over a wide range of document types, Binoculars detects over 90% of generated samples from ChatGPT (and other LLMs) at a false positive rate of 0.01%, despite not being trained on any ChatGPT data.
This paper has not been read by Pith yet.
Forward citations
Cited by 18 Pith papers
-
FLIPS: Instance-Fingerprinting for LLMs via Pseudo-random Sequences
FLIPS identifies LLM instances with 96% closed-set and 90% open-set accuracy by exploiting biases in generated binary random sequences across 237 instances.
-
SV-Detect: AI-generated Text Detection with Steering Vectors
Steering vectors from frozen LM layers enable a lightweight classifier to detect machine-generated text robustly across domains, source models, and editing attacks.
-
Detector-Evasive LLM Paraphrasing via Constrained Policy Optimization
DEPO formulates detector-evasive paraphrasing as a constrained MDP and solves it via Lagrangian primal-dual RL with GRPO-style updates to achieve evasion while satisfying a semantic-preservation constraint.
-
Base Models Look Human To AI Detectors
Base model text evades AI detectors better than instruction-tuned text, and the HIP method strengthens this trade-off across model sizes.
-
The Impact of AI-Generated Text on the Internet
By mid-2025 roughly 35% of new websites are AI-generated or AI-assisted, correlating with lower semantic diversity and higher positive sentiment but showing no significant drop in factual accuracy or stylistic diversity.
-
ExaGPT: Example-Based Machine-Generated Text Detection for Human Interpretability
ExaGPT uses span-level similarity retrieval from human and LLM datastores to detect machine-generated text while supplying the matching spans as human-interpretable evidence, achieving up to 37-point accuracy gains ov...
-
Hitting a Moving Target: Test-Time Adaptation for AI Text Detection under Continual Distribution Shift
Test-time adaptation with semi-supervised learning leverages inference-time homogeneity to maintain AI text detection performance under adversarial humanization, new LLMs, and temporal drift.
-
DetectZoo: A Unified Toolkit for AI-Generated Content Detection Across Text, Audio, and Image Modalities
DetectZoo is a unified toolkit providing reference implementations of 61 detectors, native loaders for 22 benchmark datasets, and a standardized evaluation pipeline for AI-generated content detection across text, audi...
-
Seeing the Poem: Image-Semantic Detection of AI-Generated Modern Chinese Poetry with MLLMs
An image-semantic guided method enhances MLLMs for detecting AI-generated modern Chinese poetry by combining poem text with visual representations of content, achieving 85.65% Macro-F1 with Gemini and outperforming te...
-
Steer-to-Detect: Probing Hidden Representations for Detection of LLM-Generated Texts
Steer-to-Detect learns a steering vector injected into LLM hidden states to boost class separability and applies hypothesis testing with finite-sample Type I/II error guarantees for generated-text detection.
-
RedNote-Vibe: A Dataset for Capturing Temporal Dynamics of AI-Generated Text in Lifestyle Social Media
RedNote-Vibe supplies a longitudinal dataset of AI versus human lifestyle posts from 2020 to mid-2025 plus the PLAD detection framework that applies cognitive psychology signatures for improved AI-text identification.
-
CASPER in the Machine: Insights into Character Variety in LLM-Generated Stories
LLM stories exhibit both similarities and differences from human stories in character dimensions, with LLMs producing varied characters.
-
The Assistant as a Privileged Persona: A canonical reference in cross-persona self-recognition
On Llama-3.1-70B-Instruct the Assistant persona functions as the sole canonical reference for cross-persona authorship judgments, with symmetric entropy gaps predicting only on its row and asymmetric surprise relative...
-
Paraphrasing Attack Resilience of Various AI-Generated Text Detection Methods
Binoculars-inclusive ensembles detect AI text best overall but suffer the largest performance drops under paraphrasing attacks.
-
PolyAlign: Conditional Human-Distribution Alignment
PolyAlign aligns language models to bucket-specific human response distributions via Bucket-Aware SFT and HDPO, reporting gains in conditional naturalness and distributional faithfulness on bilingual tasks.
-
From AI-Generated Content to Agentic Action: Security and Safety Threats in Generative AI
The paper analyzes evolving security and safety threats in generative AI from content generation to agentic actions, noting that attack surfaces expand faster than defenses and that many safeguards require institution...
-
Findings of the Counter Turing Test: AI-Generated Text Detection
Shared task findings show F1=1.0000 for binary AI text detection and 0.9531 for model attribution using fine-tuned DeBERTa and BART transformers with ensembles.
-
Findings of the Counter Turing Test: AI-Generated Text Detection
Shared task findings show near-perfect binary detection of AI-generated text but greater difficulty in attributing outputs to particular language models.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.