MAmmoTH2: Scaling Instructions from the Web (2024 · 96 citations)
Instruction tuning makes LLMs better at reasoning, but good instruction data is hard to get. Human annotation is costly and usually small in scale. Data distilled from GPT-4 tends to be narrow and can contain hallucinations. The authors argue that web pre-training corpora already contain plenty of question–answer…
MAP-Neo: Highly Capable and Transparent Bilingual Large Language Model Series (2024 · 44 citations)
The strongest LLMs are either closed or release only weights. They keep training data, code and intermediate checkpoints private. Fully transparent alternatives such as Pythia, Amber and OLMo share these components but trail open-weight models of similar size, especially on math, coding, knowledge and non-English…
Does Math Reasoning Improve General LLM Capabilities? Understanding Transferability of LLM Reasoning (2025 · 25 citations)
The paper asks whether the rapid gains reasoning-tuned LLMs show on math benchmarks such as MATH and AIME reflect broader problem-solving skill or narrow specialization.

Co-authors here

12 of 13 co-authors with a page, the most papers together first.

Venues

Where their works here were published, the most works first, with the editions of those that have a page.

Papers of our library cite their works 160 times (of the 149,439 papers whose bibliographies we read); 79 citing papers with a page here →.

Updated 3 November 2026