arXiv AI

Screen Before You Interpret: A Portable Validity Screen for Benchmark-Based LLM Confidence Signals

The paper introduces a portable validity screen for evaluating large language model (LLM) confidence signals, mirroring clinical personality assessment practices. It computes metrics such as the rate of high‑confidence errors (L), low‑confidence correct answers (Fp), and their contrast (RBS), without adding new statistics, to determine whether a signal truly tracks correctness. Applied to 20 frontier models on 452 two‑alternative items and 33 models on 1,500 MMLU items, the screen classifies models as Valid, Indeterminate, or Invalid, revealing that many models fail to meet validity criteria depending on confidence thresholds.

arXiv Machine Learning
Sep 25

Three Ways Classical Test Theory Misleads for LLM Judges

The article examines how classical test theory reliability statistics misrepresent the performance of large language model (LLM) judges. It shows that internal‑consistency coefficients, the dependability index, and Livingston‑Lewis accuracy each conflate judge error with item design or criterion validity, making it impossible to attribute a single reliability value to the judge alone. The authors argue that such misattribution can influence deployment decisions and documentation.

By Louis Yiven Zhu
arXiv AI
Oct 28

Scoring-Rule Sensitivity in LLM Evaluation: Evidence from Indian Financial Regulatory Text

The paper investigates how different scoring rules affect the ranking of large language models (LLMs) on the IndiaFinBench benchmark, which contains 406 expert‑annotated items covering regulatory interpretation, numerical reasoning, contradiction detection, and temporal reasoning. Using twelve LLMs evaluated under strict string‑matching, a cross‑model judge, and a composite regime, the authors find that model rankings vary dramatically across regimes, with no significant rank correlation (Spearman ρ = -0.273). Item‑level analysis reveals that a substantial portion of items are answered identically by all models, while a higher‑discrimination band isolates more challenging cases.

By Rajveer Singh Pall
arXiv AI
Sep 25

How Reproducible Are Evaluation Conclusions? A Self-Audit of LLM-Inferred Prompt Structure

The paper investigates the reliability of ranking tables produced by small-sample evaluations of large language models (LLMs). Using LLM‑inferred prompt structure across eight model variants, the authors find that prompt‑structure recovery is highly unstable, with only the bottom of the ranking consistently reproducible. They demonstrate that standard evaluation practices can misrepresent model performance and propose reporting practices to improve transparency.

By Dipankar Sarkar
arXiv Machine Learning
Sep 17

No Usable Linear "Capitulation Direction" in Two Small LLMs: A Validation Protocol for Activation-Steering Claims, and a Cross-Family Behavioral Study of Sycophancy Under Pushback

The study examines how two small instruction‑tuned language models, Qwen2.5‑1.5B and Llama‑3.2‑1B, respond to user pushback on TriviaQA. When initially correct, the models flip to a wrong answer in about 42–43% of cases, with the effectiveness of different pushback styles varying by model. Attempts to decode capitulation from the pre‑response residual stream fail under a rigorous validation protocol, revealing overfitting and a measurement hazard that underestimates capitulation by 18–24 percentage points.

By Saad Aamir, Muhammad Awais Bin Adil
arXiv AI
6d ago

HandsOnAid: Building and Evaluating an Offline First-Aid Assistant for Emergency Response Scenarios

HandsOnAid is an offline first‑aid assistant built on a 4‑bit Gemma 2B language model fine‑tuned with QLoRA on the FirstAidQA dataset. The authors evaluated seven model configurations—varying bit‑precision, safety gates, length floors, and retrieval—using a 41‑pair medically reviewed question‑answer bank and a strict 0‑5 scoring rubric with twelve safety overrides. The best performing setup, a 4‑bit fine‑tuned model with retrieval, achieved a mean score of 2.51 out of 5 and produced complete answers on a Pixel 7 Pro in under 15 seconds without network connectivity.

By Rezwan Islam Salvi, Saiyma Sittul Muna, Mushfiqur Rahman Mushfique