Pith. sign in

REVIEW 14 cited by

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2504.13171 v1 pith:KZPV76N6 submitted 2025-04-17 cs.AI cs.CL

Sleep-time Compute: Beyond Inference Scaling at Test-time

classification cs.AI cs.CL
keywords computesleep-timegsm-symbolicstatefulqueriestest-timeaimescaling
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Scaling test-time compute has emerged as a key ingredient for enabling large language models (LLMs) to solve difficult problems, but comes with high latency and inference cost. We introduce sleep-time compute, which allows models to "think" offline about contexts before queries are presented: by anticipating what queries users might ask and pre-computing useful quantities, we can significantly reduce the compute requirements at test-time. To demonstrate the efficacy of our method, we create modified versions of two reasoning tasks - Stateful GSM-Symbolic and Stateful AIME. We find that sleep-time compute can reduce the amount of test-time compute needed to achieve the same accuracy by ~ 5x on Stateful GSM-Symbolic and Stateful AIME and that by scaling sleep-time compute we can further increase accuracy by up to 13% on Stateful GSM-Symbolic and 18% on Stateful AIME. Furthermore, we introduce Multi-Query GSM-Symbolic, which extends GSM-Symbolic by including multiple related queries per context. By amortizing sleep-time compute across related queries about the same context using Multi-Query GSM-Symbolic, we can decrease the average cost per query by 2.5x. We then conduct additional analysis to understand when sleep-time compute is most effective, finding the predictability of the user query to be well correlated with the efficacy of sleep-time compute. Finally, we conduct a case-study of applying sleep-time compute to a realistic agentic SWE task.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 14 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Do Language Models Need Sleep? Offline Recurrence for Improved Online Inference

    cs.CL 2026-05 unverdicted novelty 7.0

    A sleep mechanism with N offline recurrent passes consolidates context into fast weights, improving performance on reasoning tasks where standard transformers fail.

  2. IdleSpec: Exploiting Idle Time via Speculative Planning for LLM Agents

    cs.AI 2026-05 conditional novelty 7.0

    IdleSpec improves LLM agent accuracy by generating and aggregating speculative plans during idle time between tool calls and observations using complementary drafting strategies.

  3. LongMemEval-V2: Evaluating Long-Term Agent Memory Toward Experienced Colleagues

    cs.CL 2026-05 unverdicted novelty 7.0

    LongMemEval-V2 is a new benchmark where AgentRunbook-C reaches 72.5% accuracy on long-term agent memory tasks, beating RAG baselines at 48.5% and basic coding agents at 69.3%.

  4. Post Reasoning: Improving the Performance of Non-Thinking Models at No Cost

    cs.AI 2026-05 conditional novelty 7.0

    Post-Reasoning boosts LLM accuracy by reversing the usual answer-after-reasoning order, delivering mean relative gains of 17.37% across 117 model-benchmark pairs with zero extra cost.

  5. Evaluating Memory in LLM Agents via Incremental Multi-Turn Interactions

    cs.CL 2025-07 unverdicted novelty 7.0

    MemoryAgentBench is a new multi-turn benchmark assessing four memory competencies in LLM agents—accurate retrieval, test-time learning, long-range understanding, and selective forgetting—showing that existing methods ...

  6. What to Keep, What to Forget: A Rate--Distortion View of Memory Compaction in LLMs and Agents

    cs.LG 2026-07 conditional novelty 6.0

    KV-cache eviction, prompt compression, recurrent state bounding, and agent memory consolidation are unified as one rate-distortion problem with a shared lower bound, shared failure mode, and transferable mechanisms.

  7. Playful Agentic Robot Learning

    cs.RO 2026-06 unverdicted novelty 6.0

    RATs agents generate and solve their own exploratory tasks during play, distill successful code into a skill library, and reuse it to improve held-out task performance by 20.6 and 17.0 points on two benchmarks.

  8. Language Models Need Sleep: Learning to Self-Modify and Consolidate Memories

    cs.LG 2026-06 unverdicted novelty 6.0

    Language models can use a two-stage sleep process of upward distillation for memory consolidation and RL-based dreaming for unsupervised self-improvement to enable continual learning.

  9. Anticipate and Learn: Unleashing Idle-Time Compute in Proactive Agents

    cs.CL 2026-05 unverdicted novelty 6.0

    ProAct uses idle compute to anticipate user needs via dialogue history and memory, achieving 14.8% fewer turns, 11.7% less user effort, and 28.1% fewer hallucinations than reactive baselines on the new ProActEval benchmark.

  10. Auto-Dreamer: Learning Offline Memory Consolidation for Language Agents

    cs.CL 2026-05 unverdicted novelty 6.0

    Auto-Dreamer trains an offline memory consolidator via GRPO on agent performance to abstract cross-session patterns, outperforming baselines by 7 points on ScienceWorld with 12x smaller memory and generalizing to ALFW...

  11. The Impact of Response Latency and Task Type on Human-LLM Interaction and Perception

    cs.HC 2026-02 unverdicted novelty 6.0

    Shorter LLM response latencies reduce perceived output thoughtfulness and usefulness, while task type affects prompting frequency independently of latency.

  12. Evolving Agents in the Dark: Retrospective Harness Optimization via Self-Preference

    cs.AI 2026-06 unverdicted novelty 5.0

    RHO is a self-supervised technique that selects challenging past tasks, re-solves them, and uses self-preference to update an agent's harness, raising SWE-Bench Pro pass rate from 59% to 78% without external labels.

  13. BALAR : A Bayesian Agentic Loop for Active Reasoning

    cs.AI 2026-05 unverdicted novelty 5.0

    BALAR is a task-agnostic Bayesian loop that maintains structured beliefs over latent states, selects questions via expected mutual information, and expands its state space when needed, delivering 14.6-38.5% accuracy g...

  14. Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models

    cs.CL 2025-03 accept novelty 5.0

    A survey organizing techniques to achieve efficient reasoning in LLMs by shortening chain-of-thought outputs.