paper-with-me

Papers

Incompressible Knowledge Probes: Estimating Black-Box LLM Parameter Counts via Factual Capacity

2026-04-27 · Bojie Li arxiv

Closed-source frontier labs do not disclose parameter counts. Storing F facts requires at least F/(bits per parameter) weights, so factual recall lower-bounds parameter count--an intrinsic, serving-independent signal, though (as we show) a coarse one. We introduce Incompressible Knowledge Probes (IKPs), a benchmark of 1,400 factual questions spanning 7 tiers of obscurity, designed to isolate knowledge that cannot be derived by reasoning or compressed by architectural improvements. We score with no hallucination penalty (lambda = 0: IKP accuracy is simply the fraction of probed facts answered correctly), which removes both the penalty hyperparameter and the per-tier flooring choice; a full lambda x flooring ablation shows the calibration is robust across scoring choices while individual estimates are not, motivating the no-penalty default. We calibrate a log-linear mapping from IKP accuracy to parameter count on 93 open-weight models (135M-1,600B) spanning 19 vendors, achieving R^2 = 0.910; leave-one-out cross-validation confirms generalization (median fold error 1.48x, 72% within 2x, 86% within 3x). The instrument is deliberately coarse--its 90% prediction interval spans ~3x in either direction, wider than inference economics--so IKP recovers order-of-magnitude effective capacity and relative rankings, not precise parameter counts. For Mixture-of-Experts models, total parameters predict knowledge (R^2 = 0.67) better than active parameters (R^2 = 0.41). We evaluate 201 models from 27 vendors on a curated probe set (1,311 of 1,400 probes surviving name-collision and label-ambiguity filters) and report effective knowledge capacity for all major proprietary frontier models as prediction bands rather than point estimates; for heavily safety-tuned models these are lower bounds, since refusal policy can suppress tens of percentage points of otherwise-answerable capacity.

📄 PDF Abstract BibTeX arXiv:2604.24827

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Benchmarking Deception Probes via Black-to-White Performance Boosts

2025-07-16 · Avi Parrack, Carlo Leonardo Attubato, Stefan Heimersheim arxiv

AI assistants will occasionally respond deceptively to user queries. Recently, linear classifiers (called "deception probes") have been trained to distinguish the internal activations of a language model during deceptive…

No Answer Needed: Predicting LLM Answer Accuracy from Question-Only Linear Probes

2025-09-12 · Iván Vicente Moreno Cencerrado, Arnau Padrés Masdemont, Anton Gonzalvez Hawthorne, David Demitri Africa 외 arxiv

Do large language models (LLMs) anticipate when they will answer correctly? To study this, we extract activations after a question is read but before any tokens are generated, and train linear probes to predict whether t…

Mathematical Reasoning

Linear Probe Accuracy Scales with Model Size and Benefits from Multi-Layer Ensembling

2026-04-15 · Erik Nordby, Tasha Pais, Aviel Parrack arxiv

Linear probes can detect when language models produce outputs they "know" are wrong, a capability relevant to both deception and reward hacking. However, single-layer probes are fragile: the best layer varies across mode…

BiG-SURE - Bipartite Graph for Semantic Uncertainty and Reliability Estimation of LLMs

2026-08-31 · Debarpan Bhattacharya, Malay Phadke, Sriram Ganapathy arxiv

Reliable uncertainty estimation is a crucial requirement for deploying large language models (LLMs) and vision-language models (VLMs) in safety-critical settings, especially when the model parameters are not accessible (…

Red-teaming Activation Probes using Prompted LLMs

2025-11-01 · Phil Blandfort, Robert Graham arxiv

Activation probes are attractive monitors for AI systems due to low cost and latency, but their real-world robustness remains underexplored. We ask: What failure modes arise under realistic, black-box adversarial pressur…