Rishabh Tiwari

Restricting LLM fine-tuning to the bottom singular subspace nearly eliminates catastrophic forgetting, even when protecting 99% of weights, while maintaining strong target-task performance.

Asynchronous OPD can boost training throughput significantly while managing the challenges of stale data, transforming the efficiency of large language model fine-tuning.

Verifier-free evolution can now match or exceed the performance of verifier-based methods, while slashing API costs by 3x and boosting throughput by 10x, thanks to a clever model orchestration strategy.

Models are substantially better at pairwise self-verification than independent scoring, unlocking a more efficient and accurate approach to test-time scaling for complex reasoning.