25 results

Preprint

A Detector Score Is Not a Release Risk: What PII Detection Benchmarks Certify About Deployment, Why 'Rivals Human Annotators' and 'Fails Under Shift' Are Both Measured, and the Coverage Audit a Release Decision Needs

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research Automatic detection of personally identifiable information (PII) is the step that decides what text may leave an organisation, and its published record appears to contradict itself. Clinical de-identification systems have been…

Cite Related research LOR-00174 · 2 views · 2 downloads Select “A Detector Score Is Not a Release Risk: What PII Detection…” for export

PDF357.2 KB

Preprint

A Held-Out Site Is a Sample of One: What Random, Spatial and Site-Disjoint Splits Each Estimate, Why Leakage That Differs Between Configurations Reorders a Comparison, and How Many Sites a General Claim Needs

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research A September 2026 study of farm-machinery traversability reports that a forward-facing near-infrared camera appears to beat a colour camera on a paddy-boundary class under its corpus's own sequence-level split, that the…

Cite Related research LOR-00170 · 3 views · 6 downloads Select “A Held-Out Site Is a Sample of One: What Random, Spatial…” for export

PDF380.7 KB

Preprint

A Domain Label Is Not a Treatment: Why Specialist and General-Purpose Language Models Trade Places Across Studies, What a 'Medical' Checkpoint Leaves Unspecified, and the Matched Contrasts That Would Assign the Gain

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research A September 2026 study compares an 8B general-purpose instruction-tuned model with a medically fine-tuned checkpoint built on the same base and finds that the general model scores higher both at selecting the meaning of…

Cite Related research LOR-00162 · 5 views · 12 downloads Select “A Domain Label Is Not a Treatment: Why Specialist and…” for export

PDF405.1 KB

Preprint

Form Rescued Is Not Content Gained: What Constrained Decoding Does to Accuracy, Why Published Studies Disagree About Its Sign, and the Parse-Failure Accounting That Decides the Small-Model Case

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research Constrained decoding masks every token that would break a grammar or a JSON schema, so the output always parses. Whether it also changes how often the answer is right is contested: one line of work reports that format…

Cite Related research LOR-00149 · 8 views · 9 downloads Select “Form Rescued Is Not Content Gained: What Constrained…” for export

PDF383.3 KB

Preprint

The Incumbent Did the Work: What the Simple-Baseline Results in LLM-Guided Program Search Settle, Why the Dispute Over Archives, Diversity and Test-Time Training Is Fought on Benchmarks That Cannot Rank It, and the Comparisons That Would

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research Systems that let a language model write programs, score them with an exact verifier and feed the best ones back have improved published bounds on open mathematical problems, and each new system has added machinery: program…

Cite Related research LOR-00148 · 10 views · 7 downloads Select “The Incumbent Did the Work: What the Simple-Baseline…” for export

PDF392.1 KB

Preprint

Which Fine-Tuning Failed? The Reported Failure and Recovery of Fine-Tuning in Model Editing Are Findings About the Edit Protocol, Why the Breadth-First Recovery Changes What the Editor May Remember, and the Matched Comparisons That Would Assign the Credit

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research Fine-tuning has been treated as the weak baseline of model editing since the field's first benchmarks. A 2025 study argues that the weakness was an artefact of implementation: when edits are optimised in epochs over…

Cite Related research LOR-00141 · 9 views · 8 downloads Select “Which Fine-Tuning Failed? The Reported Failure and…” for export

PDF373.9 KB

Preprint

What a Word Limit Removes: Length Constraints Look Free, Costly or Helpful Depending on the Span They Bind, the Scorer That Reads the Answer and the Items Chosen, and Why Compliance Looks Orthogonal to Accuracy in Only One Kind of Design

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research Published work on telling a language model to be brief disagrees about almost everything. One benchmark reports that obeying a word limit and answering accurately are statistically unrelated, and that alignment for helpfulness…

Cite Related research LOR-00136 · 9 views · 9 downloads Select “What a Word Limit Removes: Length Constraints Look Free…” for export

PDF364.1 KB

Preprint

Calibrated Is Not Informative: What a Fall in Expected Calibration Error Shows About LLM Confidence, Why Verbalised Scores Fail on Resolution as Well as Level, and the Recalibration Baseline the Ensembling Results Omit

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research A common finding in work on language-model confidence is that a model asked to state its own confidence in one pass is badly overconfident, and that some added procedure - sampling several answers, a separate critique pass, a…

Cite Related research LOR-00133 · 14 views · 10 downloads Select “Calibrated Is Not Informative: What a Fall in Expected…” for export

PDF352.5 KB

Preprint

Tiny in Parameters, Not in Compute: What a Recursive Reasoning Model's Puzzle Score Credits to Recursion, Why the Ablations Split by Task, and the Matched Comparisons That Would Assign the Rest

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research Recursive reasoning models, which apply one small network many times to a latent state, report accuracies on Sudoku, maze and ARC-AGI puzzles above those of most frontier language models in their comparison tables, with 5 to…

Cite Related research LOR-00132 · 14 views · 9 downloads Select “Tiny in Parameters, Not in Compute: What a Recursive…” for export

PDF391.1 KB

Preprint

A Starting Score Is Not a Ceiling: When Accuracy Before Reinforcement Learning Predicts Accuracy After It, Why Each Proposed Replacement Sees a Different Failure, and the Axis Test That Decides Which Proxy to Trust

Pranay Mahendrakar – 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research Reasoning models are usually built in two stages, supervised fine-tuning and then reinforcement learning against a verifier, and the first stage is usually judged by the benchmark accuracy of its checkpoint. An August 2026…

Cite Related research LOR-00129 · 11 views · 11 downloads Select “A Starting Score Is Not a Ceiling: When Accuracy Before…” for export

PDF326 KB