21 / papers

Safety & Verification

Guarantees, specifications, and the gap between a proof and a deployed system. 21 open-access papers by Pranay Mahendrakar, each with a permanent DOI.

preprint

A Null Citation Gap Is Not a Missing Reward: What Field-Level Comparisons of Papers With and Without Shared Data and Code Certify About the Incentive to Share, Why a Transportation Null and an Astrophysics Advantage Can Both Hold, and the Design That Would Price the Return

Life of Research Whether sharing data and code earns authors more citations is now contested in print. A 2026 study that used a large language model to label 10,724 transportation papers reports no citation difference for either practice…

preprint

A Relocated Dependency Is Not a Vanished One: What the Neural-Cellular-Automaton Result on Integer Multiplication Shows About Long-Range Dependency, Why the Neural GPU Line Already Bounds It, and the Controls That Would Separate Locality, Supervision and Carry Coverage

Life of Research Multi-digit multiplication is a standing failure case for neural sequence models, and a 2025 reverse-engineering study traced the failure of standard fine-tuning to long-range dependencies that the model never learns. A…

preprint

A Missing Slice Is a Missing Measurement: What Language-Level Audits of Safety Datasets Establish, Why the Multi-Turn Residual and Cross-Lingual Transfer Break the Step From Data Gaps to Model Behaviour, and the Contrasts That Would Price Native Data

Life of Research Model providers cite multilingual safety benchmarks as evidence that their models are safe for speakers of many languages, and a 2026 audit shows that such claims often dissolve when a single language is inspected: across…

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

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…

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

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…

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

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…

preprint

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

Life of Research (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…

preprint

Robust Because Coarse, or Robust Because Planned? A Research Agenda for Training-Data Provenance After the Aggregate-Disclosure Turn, in Which Granularity Buys Statistical Power, Prior Randomisation Buys Soundness, and Per-Example Value Has Neither

Life of Research (c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Regulators have chosen coarse disclosure of training data: the template for the EU AI Act's public summary of training content asks for lists of large datasets, a…

preprint

INT4 Is Not One Treatment: Why the Quantisation-Safety Literature Cannot Be Reconciled as Reported, How Six Design Axes Separate Its Studies Before Bit-Width Does, and Why Two 2026 Studies of the Same Model Names Reach Opposite Verdicts

Life of Research (c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Post-training quantisation to 4 bits is widely treated as nearly free, a judgement resting on perplexity and multiple-choice accuracy. Whether it is also free for…

preprint

Where the Transcript Runs Out: When Model Internals Beat Behavioural Monitors, Why the Strongest Generalisation Evidence Comes From Targets the Text Already Shows, and the Comparisons No Study Has Yet Run

Life of Research (c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Linear probes and other readers of a language model's activations are now deployed as safety monitors, and two published literatures disagree about what they buy. One…

preprint

One Pass, Three Purchases: Deduplication's Efficiency, Generalisation and Privacy Arguments Each Hold in Their Own Regime, Rest on Different Properties of the Repetition Distribution, and Cannot All Be Bought With One Similarity Threshold

Life of Research (c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Deduplication became a fixed stage of language-model pretraining on the strength of three claims made together: removing duplicates saves compute, improves what the…

preprint

A Stated Limit Is a Forecast: What an Agent Can Truthfully Say About Its Own Permissions, Why the Checkable Part Should Never Come From the Model, and How to Score the Part Only the Model Can Supply

Life of Research (c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Agent interfaces, skill specifications and disclosure proposals increasingly ask a tool-using language-model agent to tell its user what it may and may not do. The…

preprint

The Precursor Assumption: What an Early-Warning Threshold Must Deliver in a Frontier Safety Framework, Why the Continuity Evidence Covers Aggregates Under Fixed Elicitation Rather Than the Thresholded Task, and the Lead Time No Published Record Measures

Life of Research (c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Frontier safety frameworks decide when a model needs stronger safeguards by testing it against capability thresholds at a fixed cadence, and they place an…

preprint

Fingerprints That Survive What? Model Lineage Names Three Relations, Robustness Is Indexed by Which Party Is Adversarial, and a Benchmark's Distillation Column Scores the Parent That Was Not Distilled From

Life of Research (c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Language-model fingerprinting is asked to answer questions of the form "is this model derived from that one?", and its methods are routinely reported as robust to…

About this research line

What has Pranay Mahendrakar published on Safety & Verification?

Pranay Mahendrakar has published 21 open-access papers on Safety & Verification: "A Null Citation Gap Is Not a Missing Reward", "A Relocated Dependency Is Not a Vanished One", "A Missing Slice Is a Missing Measurement", "Which Way the Verdict Falls", "A Tally of Framings Is Not a Census of Causes", "A Named Answer Is Not a Correction", "Hard to Read Is Not Hard to Audit", "Robust Because Coarse, or Robust Because Planned? A Research Agenda for Training-Data Provenance After the Aggregate-Disclosure Turn, in Which Granularity Buys Statistical Power, Prior Randomisation Buys Soundness, and Per-Example Value Has Neither", "INT4 Is Not One Treatment", "Where the Transcript Runs Out", "A Control Knob Is Not a Concept", "One Pass, Three Purchases", "A Stated Limit Is a Forecast", "The Precursor Assumption", "Fingerprints That Survive What? Model Lineage Names Three Relations, Robustness Is Indexed by Which Party Is Adversarial, and a Benchmark's Distillation Column Scores the Parent That Was Not Distilled From", "Refusal Is Not a Rate", "Checked at Every Step Is Not Checked as a Whole", "Who Pulls the Plug? Self-Report, External Authority, and the Limits of Self-Isolation in AI Control", "Formal Verification of Neural Network Safety Beyond Toy Examples", "AI-Generated Text Detection Under Paraphrasing" and "Emergent Covert Signaling in Multi-Agent LLM Negotiation". Each is deposited on Zenodo with a permanent DOI and is free to read.

← All research topics