The paper investigates how test‑time computation can enhance language models and at what cost, introducing the SELF‑POT benchmark to evaluate this across competition mathematics, competitive programming, and agentic workflows. SELF‑POT separates candidate coverage from final accuracy, tracks correctness transitions under revision, and measures protocol completion alongside task success. Using a unified budget rule, the study compares direct inference, parallel sampling, and self‑revision across five low‑cost reasoning models, revealing that selection rules and failure handling significantly influence gains and cost savings.
arXiv:2609.15309v1 Announce Type: new Abstract: Large language model (LLM) agents allocate test-time compute adaptively as they revise solutions, use tools, explore alternatives, and decide when to s...
arXiv:2606. 04402v1 Announce Type: new Abstract: Modern reasoning models can allocate different amounts of test-time computation, such as thinking tokens, model calls, or compute budget, to different tasks.
arXiv:2606. 01667v1 Announce Type: new Abstract: Test-time scaling has become a major way to improve large language model reasoning, but its orchestration has remained designer-engineered: a fixed sample budget, a fixed refinement loop, a fixed scoring rule, or a fixed search policy decides how compute is spent, leaving the model in charge of solving but not of orchestration.
The paper demonstrates that the number of candidates generated during test-time scaling of large language models does not fully capture the system cost. By comparing different generation schedules (e.g., one batched call versus multiple serial calls) while keeping the total candidate count fixed, the authors show that serial calls consume significantly more GPU energy and latency. The study suggests that reporting candidate count alone is insufficient; evaluations should also include generation schedule and GPU-level metrics.
The paper proposes a unified framework for test‑time reasoning methods, framing them as recursion operators—GROW, PRUNE, and BRANCH—applied to an agent’s reasoning trace. Experiments across five benchmarks and three frontier models show that BRANCH, which samples and selects among multiple reasoning paths, consistently outperforms the other operators and a single‑pass chain‑of‑thought baseline, improving accuracy by an average of 5.98 percentage points. The study also highlights the importance of paired evaluation and careful handling of scoring‑pipeline failures, as these factors can significantly alter comparative outcomes.
Test-time scaling has become a major way to improve large language model reasoning, but its orchestration has remained designer-engineered: a fixed sample budget, a fixed refinement loop, a fixed scoring rule, or a fixed search policy decides how compute is spent, leaving the model in charge of solving but not of orchestration. We introduce ATLAS, an agentic test-time scaling framework in which an LLM orchestrator owns the control loop end-to-end.
The paper introduces agentic meta‑reasoning, a structured inference‑time framework that explicitly manages control decisions—such as selecting partial work, restarting, or stopping—during long‑horizon agentic tasks. By delegating task execution to workers and consolidating decisions through a lightweight controller that references persistent memory, the method reduces the need to replay full histories. Experiments on ProgramBench and other benchmarks show that meta‑reasoning improves performance over direct control baselines, especially as computation budgets increase, and reveals greater reuse of earlier work and higher solution coverage.
arXiv:2609.38612v1 Announce Type: new Abstract: As natural language drives more applications, language models increasingly run inside programs as decision components: the program sends them the curre...
In cognitive science, resource rationality asks how an agent should allocate limited computation to maximize expected value. Most reasoning and agent benchmarks use independent per-task budgets; exist...
arXiv:2601. 21522v2 Announce Type: replace-cross Abstract: The performance of large language models (LLMs) on verifiable tasks is usually measured by pass@k, the probability of answering a question correctly at least once in k trials.
arXiv:2608. 12307v1 Announce Type: cross Abstract: Recent work on distillation transfers the capabilities of large models to smaller ones often by updating the latter's parameters, through teacher forcing, on-policy distillation, and related training-time methods.
The paper investigates how reinforcement learning post‑training of large language models (LLMs) tends to sharpen existing behaviors, improving single‑shot accuracy but reducing solution coverage. It shows that pre‑trained LLMs, when paired with a lightweight inference harness, can outperform post‑trained models in coverage for agentic tasks that require multi‑turn tool use. The authors introduce the Sharpening Tax metric to quantify this trade‑off, analyze its prevalence across 14 model pairs and 42 benchmark cases, and propose a Bayesian sampler, PTGS, that mitigates the tax by adapting sampling temperature to prompt difficulty.