Compute-Optimal Test-Time Scaling

Sample Count Is Not Enough: Candidate-Generation Strategy Shapes the Energy and Performance of LLM Test-Time Scaling
arXiv:2609.19499 — For self-consistency, RASC, ESC, and best-of-N, candidate count N is not enough. Batching the same N can change A100 energy and latency by several times.
Mobina Kashaniyan, Ali Jannesari
Posted Nov 1, 2026 Conference Paper, Workshop Paper, Preprint, Large Language Models, High-Performance Computing, Energy Efficiency
PDF DOI PDF arXiv DOI Google Scholar Research Summary Local PDF
ATEST / RATTS: Adaptive Resource-Aware Test-Time Scaling for LLM Inference (arXiv:2610.25519)
ATEST (RATTS) makes self-consistency resource-aware, unlike fixed self-consistency, RASC, ESC, and Adaptive-Consistency. The same four-sample query can use 4.1× more energy on a busy GPU. arXiv:2610.25519.
Mobina Kashaniyan
Posted Oct 18, 2026 Large Language Models, High-Performance Computing, Energy Efficiency, Test-Time Scaling, Research Summary
ATEST: Adaptive Resource-Aware Test-Time Scaling for LLM Inference
ATEST (RATTS) is resource-aware self-consistency. Unlike fixed self-consistency, RASC, ESC, and Adaptive-Consistency, it also uses live GPU cost. arXiv:2610.25519.
Mobina Kashaniyan, Ali Jannesari
Posted Oct 18, 2026 Preprint, Large Language Models, High-Performance Computing, Energy Efficiency, Test-Time Scaling
PDF DOI PDF arXiv DOI Google Scholar Research Summary Local PDF
Sample Count Is Not Enough (arXiv:2609.19499): Why Candidate-Generation Strategy Matters for LLM Test-Time Scaling Energy and Performance
arXiv:2609.19499 — For self-consistency, RASC, ESC, Universal Self-Consistency, and best-of-N, sample count N is not enough. At N=8, 1×8 vs 8×1 changes A100 energy by about 4.6–4.9×.
Mobina Kashaniyan
Posted Sep 16, 2026 Last updated Sep 17, 2026 Large Language Models, High-Performance Computing, Energy Efficiency, Test-Time Scaling, Research Summary
Interpretable Adaptive Sampling for LLM Test-Time Scaling
Interpretable adaptive self-consistency. A fuzzy controller sets the sample budget, compared with best-of-N, self-certainty, majority voting, and compute-optimal test-time scaling.
Mobina Kashaniyan, Ali Jannesari
Posted Aug 4, 2026 Preprint, Artificial Intelligence, Large Language Models
PDF DOI PDF arXiv Google Scholar Research Summary
Interpretable Adaptive Sampling for LLM Test-Time Scaling and Test-Time Compute
Interpretable adaptive self-consistency versus best-of-N, self-certainty, majority voting, and compute-optimal test-time scaling. Related methods include RASC, ESC, and Adaptive-Consistency.
Mobina Kashaniyan
Posted Aug 4, 2026 Large Language Models, Artificial Intelligence, LLM Reasoning, Test-Time Scaling, Research Summary

© 2026 Mobina Kashaniyan.