Mohit Iyyer · page 2

29 papers in our library · 2014–2026 · page 2 of 2

Papers

The most cited first (OpenAlex and our library).

Exploring the impact of low-rank adaptation on the performance, efficiency, and regularization of RLHF (2023 · 9 citations)
The final stage of RLHF, PPO training, is very memory-hungry. The policy, value, reward and reference models must all sit in GPU memory together. With the AlpacaFarm library, aligning LLaMA 7B this way takes eight A100 80GB GPUs. The authors adapt AlpacaFarm so that PPO updates only small LoRA adapters (rank 8, α =…
AI use in American newspapers is widespread, uneven, and rarely disclosed (2026 · 7 citations)
Newsrooms are adopting AI tools quickly, but no one has measured how much machine-generated text actually reaches readers of American newspapers.
Frankentext: Stitching random text fragments into long-form narratives (2026 · 1 citation)
The paper asks whether a black-box LLM user can cheaply produce long stories that read well and still slip past strong AI-text detectors.
AnalystBench: Benchmarking professional long-form report generation with web-mined multimodal tasks (2026)
Benchmarks for LLM-based report writing usually test isolated subskills, such as summarization or table aggregation. Benchmarks with realistic professional tasks, like GDPval, depend on costly expert-built tasks. AnalystBench aims to evaluate the whole analyst workflow at low cost.