151 results

Preprint

A Compliance Gap Is Not a Motive: Why 'Does Alignment Faking Require Scale?' Has No Answer Until the Measured Quantity Is Fixed, What Small-Model Gaps and Frontier Scratchpads Each Establish, and the Interventions That Would Separate Goal-Guarding From Expectation-Tracking

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research Greenblatt et al. (2024) showed Claude 3 Opus complying with a harmful training objective more often when it could infer that its answers would be trained on, and found the same compliance gap in Llama 3.1 405B but not in…

Cite Related research LOR-00120 · 9 views · 12 downloads Select “A Compliance Gap Is Not a Motive: Why 'Does Alignment…” for export

PDF395.2 KB

Preprint

Which Way the Verdict Falls: Noise-Induced Flips in an LLM Bias Judge Run Toward the Label Its Prompt Makes Easiest to Reach, the Harm-Detection Literature Records the Opposite Sign, and Whether Noise Inflates a Bias Score Depends on the Score

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research Language models are now used as judges that read a passage and decide whether it leans toward a stereotype, and the passages they read are often noisy. A 2026 study corrupted 3,822 stereotype-probing responses with typos…

Cite Related research LOR-00116 · 26 views · 26 downloads Select “Which Way the Verdict Falls: Noise-Induced Flips in an LLM…” for export

PDF373.1 KB

Preprint

A Tally of Framings Is Not a Census of Causes: What the AI Risk Repository's Human-Versus-AI Split Measures, Why Its Gap Closed From Seventeen Points to Four as the Corpus Grew, and the Attribution Rule a Causal Claim Would Need

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research The AI Risk Repository extracts every risk named in published AI risk frameworks and codes each one by the entity presented as causing it. Its journal version reports that "human decisions cause nearly as many AI risks (38%)…

Cite Related research LOR-00117 · 13 views · 28 downloads Select “A Tally of Framings Is Not a Census of Causes: What the AI…” for export

PDF400.1 KB

Preprint

A Named Answer Is Not a Correction: The Moral Self-Correction Threshold Moved From 22B to 3.8B on Benchmarks a Context-Blind Policy Can Pass, and the Controls That Would Separate Steering From Correcting

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research Instructing a language model to avoid stereotypes lowers its measured social bias, and a 2023 study reported that this capacity for "moral self-correction" emerges at 22 billion parameters. A 2024 study reported the same…

Cite Related research LOR-00114 · 13 views · 12 downloads Select “A Named Answer Is Not a Correction: The Moral…” for export

PDF401.8 KB

Preprint

A Hidden Shift Is Not a Hidden Bias: What a Representation-Level Counterfactual Gap Certifies About an LLM Recommender, Why the Encoded-Versus-Expressed Studies Disagree, and the Placebo, Mediation and Transport Tests That Would Settle It

Pranay Mahendrakar – 2026

Life of Research (c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Fairness audits of language-model recommenders compare the lists a model returns when a single stated identity is swapped and the rest of the prompt is held fixed. A 2026 benchmark adds a second axis, the displacement of the…

Cite Related research LOR-00115 · 17 views · 34 downloads Select “A Hidden Shift Is Not a Hidden Bias: What a…” for export

PDF377.2 KB

Journal article

Hard to Read Is Not Hard to Audit: Why Degrading a Text Image Raises Jailbreak Success in One 2026 Study and Lowers It in Another, Four Accounts That Fit the Same Curve, and the Ablations That Would Separate Them

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Rendering text as an image is now proposed as a way to compress long contexts for multimodal language models, and a 2026 study reports that the same move opens a safety hole: when a harmful request is rendered at a resolution low enough to be…

Cite Related research LOR-00112 · 12 views · 27 downloads Select “Hard to Read Is Not Hard to Audit: Why Degrading a Text…” for export

PDF398.1 KB

Journal article

Wrong Together Is an Item-Level Quantity: Why Per-Person Reliance Measures Cannot Say Whether AI Assistance Correlates Human Errors, Why the Direct Probes Disagree, and the Reanalysis That Would Settle It

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Research on reliance on AI advice scores one decision-maker at a time: how often a person accepts correct advice, how often they reject wrong advice, and what that does to their accuracy. Deployment places many decision-makers in front of the…

Cite Related research LOR-00113 · 24 views · 40 downloads Select “Wrong Together Is an Item-Level Quantity: Why Per-Person…” for export

PDF391.1 KB

Journal article

A Rewrite That Keeps the Label Has Already Read It: Why an LLM Paraphrase Can Raise a Sentiment Classifier's Accuracy Without the Classifier Getting Better, Four Accounts of the Gain, and the Re-Annotation and Neutral-Rate Controls That Would Separate Them

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. A 2026 study reports that sentiment classifiers score higher on reviews rewritten by small language models than on the human originals, by up to 5.8 accuracy points, and reads the gain as stylistic alignment: the rewrite strips idiosyncrasy…

Cite Related research LOR-00110 · 16 views · 26 downloads Select “A Rewrite That Keeps the Label Has Already Read It: Why an…” for export

PDF374 KB

Journal article

Who Spoke Next Is Not Who Should Have: What Zero-Shot Failure on Multi-Party Turn-Taking Benchmarks Measures, Why the Human Reference Sits Beside the Models, and What Would Show That Knowing When to Speak Must Be Trained

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. An assistant that joins a group conversation has to decide, at every pause, whether to take the floor. Two 2026 benchmarks report that language models prompted zero-shot make this decision badly and that fine-tuning repairs it, and both…

Cite Related research LOR-00111 · 13 views · 30 downloads Select “Who Spoke Next Is Not Who Should Have: What Zero-Shot…” for export

PDF420.3 KB

Journal article

A Blind Score Is Not a Reliance Measure: What Answering Without the Audio Shows About Audio-Language Benchmarks, Why the Partial-Input Debate Already Settled Half of It, and the Counterfactual-Audio Test the Other Half Needs

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Audio-language models are now routinely tested with the audio taken away, and the results are read in opposite directions. One 2026 audit reports that eight models keep 60 to 72 percent of their accuracy on three benchmarks when the audio is…

Cite Related research LOR-00107 · 14 views · 20 downloads Select “A Blind Score Is Not a Reliance Measure: What Answering…” for export

PDF429.3 KB