Pith. sign in

REVIEW 5 major objections 4 minor 1 cited by

Vulcan: Instance-specialized, Verifiable Systems Heuristics Through LLM-driven Search

T0 review · 5 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that systems heuristics can be synthesized per deployment instance by having LLMs evolve small stateless scoring functions against trusted mechanisms, and that the resulting policies match or beat hand-designed state-of-th

desk verdict The VALUE/RANK interface is a genuinely useful reframing, but the empirical claims are not supported as stated due to train/test leakage and an abstract that overstates the body's own results; worth reading for the idea, not yet for the numbers. read the letter →

arxiv 2512.25065 v2 pith:Z5TAER47 submitted 2025-12-31 cs.OS cs.AIcs.DC

Rohit Dwivedula , Divyanshu Saxena , Sujay Yadalam , Eric Hayden Campbell , Daehyeok Kim , Aditya Akella This is my paper
classification cs.OScs.AIcs.DC
keywords systemsheuristicsLLMcodesynthesisinstance-optimalpoliciescacheevictionmemorytieringevolutionarysearchpolicy-mechanismseparationVALUE/RANKinterfaces
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that systems heuristics can be synthesized rather than hand-authored, and that the reason LLM-based synthesis fails in this domain is structural: current heuristics entangle decision logic with low-level mechanisms. Vulcan's answer is to separate the two with narrow interfaces—VALUE, a function from state to a scalar decision, and RANK, a per-object scoring function—and to let evolutionary search over LLM-generated scoring functions find instance-specialized policies. The paper reports that on ten cache-eviction workload clusters the synthesized policies match or beat state-of-the-art baselines, by up to 69% in one cluster, and that queue-topology variants beat seventeen baselines by 1.0–3.2%; in memory tiering, synthesized promotion policies beat the ARMS system by 2.5–7.9% across four application instances. The broader aspiration is to change the cost model of heuristic design so that specialization becomes the default, with humans defining interfaces and evaluation harnesses instead of tuning constants.

What carries the argument

The VALUE and RANK interfaces. VALUE reduces a policy to a function value(X) computing a scalar from system features (e.g., cwnd); RANK reduces it to a per-object score(X, o_i) whose top-K selection is performed by a reusable mechanism (full sort, sample sort, or priority queue). The evolution loop pairs an LLM generator with an evaluator harness; the template constrains the function signature and features, while the harness returns a single optimization metric. A second, 'queue topology' form asks the LLM to co-design initial-placement and transition functions among a small set of FIFO/LRU queues—a value-style coding of routing decisions—which yields constant-time eviction policies. The cen

What would settle it

Take one trace per cluster, run Vulcan's search to completion, then measure the discovered policy on held-out traces in the same cluster as well as on traces from other clusters. If the per-cluster advantage over GDSF or S3-FIFO vanishes or reverses, the instance-specialization claim is not supported; if it survives, the claim is robust. A second check: compare the best policy selected on the search trace against a policy found by random search with the same number of evaluations.

Watch

Extended reading notes

Core claim

The central claim is that constraining an LLM to write a single stateless scoring function—rather than an entire mechanism-entangled heuristic—turns heuristic synthesis into a tractable search problem, and that the resulting policies can beat hand-designed ones for a specific deployment instance. The interfaces are the load-bearing invention: every task is recast as either value(X), computing a scalar, or score(X, o_i), ranking objects, and the mechanism (priority queue, full sort, queue topology) is provided by trusted scaffolding. The paper claims this makes validation nearly trivial: any real-valued function is a well-formed policy, so 'it may be a poor policy, but it cannot be an invalid

Load-bearing premise

The load-bearing premise is that one trace drawn from a cluster is representative enough of that cluster to serve as the search objective, while the final reported cluster average includes that same trace; if that trace is not representative, the reported gains may be partly an artifact of selection rather than specialization.

Editorial extensions

If this is right

  • If the claim holds, heuristic redesign stops being a human bottleneck: the same template can be pointed at a new workload cluster and, in hours, produce a specialized policy whose cost is measured in API calls rather than engineer months.
  • Interface safety means synthesized policies can be put on the hot path without a separate verification layer; a function that returns a number cannot break the system even if it is stupid.
  • The interface taxonomy predicts which tasks can be automated: the paper's survey of 660 recent systems papers finds 71 VALUE tasks and 158 RANK tasks among 234 identified resource-management tasks, so the method should transfer to scheduling, prefetching, congestion control, and admission control.
  • Instance-specialization can become continuous: an automated instance classifier can notice a workload shift, trigger a new search, and deploy a new heuristic, making one-size-fits-all a design choice rather than a necessity.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper: the learned scoring functions double as a readable explanation of what matters for an instance (e.g., NVM bandwidth saturation for GUPS, burst-phase detection for Silo), so the same pipeline could be used as an automated workload-characterization tool.
  • Beyond the paper: because the search uses one trace per cluster during selection and reports cluster averages that include that trace, a held-out evaluation would be needed to confirm that the discovered heuristics generalize rather than overfit; this is an open question the paper leaves implicit.
  • Beyond the paper: the abstract advertises spot-VM scheduling savings, but the body's evaluation covers cache eviction and memory tiering only; transferring the interface to admission control would require building the mechanism and harness for that domain.
  • Beyond the paper: the low search cost suggests continuous re-specialization is feasible—an instance classifier could trigger a fresh search whenever the workload drifts, making the heuristic itself a managed resource.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 4 minor

Summary. The paper proposes VULCAN, a framework that uses LLM-driven evolutionary search to synthesize instance-specialized systems heuristics. The key idea is to separate policy from mechanism through VALUE and RANK interfaces, so that the LLM is only asked to generate a stateless scoring or ranking function while trusted scaffolding handles the rest. The authors instantiate the framework on cache eviction and memory tiering, and report that synthesized cache policies outperform strong baselines by up to 69% in per-cluster miss-rate reduction, and that tiering policies improve on vanilla ARMS by 2.5-7.9%. The paper also presents an LLM-assisted survey of 660 OSDI/NSDI papers to argue that the VALUE/RANK interfaces are broadly applicable.

Significance. If validated, the paper would make a useful contribution to automated systems heuristic design: the interface abstraction is clean, the generated heuristics are human-readable, and the idea of specializing policies to instances is timely. The paper also provides code pointers ([27], OpenEvolve [90]) and a large-scale LLM-assisted survey of the literature in Appendix A, which is a useful auxiliary contribution. However, the empirical evaluation as presented does not establish the central claims: the cache evaluation has a train/test contamination problem, the abstract advertises contributions that do not appear in the body, and the memory-tiering results rest on a narrow comparison without variance analysis.

major comments (5)
  1. [§4.1.3–§4.1.4] The cache-evaluation protocol is contaminated. The text states that "the evaluator harness uses one trace from within the cluster to score candidate solutions," and that the final heuristic "is then evaluated on all traces within this cluster." Thus the cluster-averaged MRR reported in Figure 8 includes the very trace used to select the heuristic. The identity of the search trace is not disclosed, no held-out split is performed, and no per-trace results are reported. The advertised gains (1.94–69%, including the 69% result for C2) are therefore not a valid estimate of generalization; they may be inflated by overfitting to a single trace. A held-out evaluation, with the search trace excluded and per-trace results reported, is required before the main cache claims can be assessed.
  2. [Abstract vs. body] The abstract advertises support for spot-VM scheduling ("up to 4.9x higher savings") and a restricted language called Anvil that "guarantees important properties by construction." Neither spot-VM nor Anvil is defined or evaluated anywhere in the body. The abstract's headline numbers also do not match the body: the abstract says "up to 2x lower miss ratios" and "up to 10% higher application performance," while the body reports 1.94–69% MRR for cache eviction and 2.5–7.9% for memory tiering. This makes it unclear which claims are actually being defended and overstates the evaluated scope of the work.
  3. [§4.1.4, Figure 8] The text itself reports that the synthesized heuristic is best in only three of ten clusters (C1, C2, C3), is second to GDSF in four clusters (C5, C6, C8, C9), and is third in the remaining clusters. This contradicts the full-text abstract's statement that the heuristics "outperform all human-designed state-of-the-art algorithms." It also weakens the central claim of instance-optimality: in seven of ten instances the synthesized policy ranks second or third. Figure 8 is presented without per-cluster numeric values or error bars, making the magnitude of the advantage impossible to evaluate.
  4. [§4.2.4] The queue-topology results for C7 and C8 use the same instance-generation protocol as §4.1, so the same train/test contamination concern applies: candidate topologies are scored on a trace from the cluster and then evaluated on the cluster, with no explicit held-out split. In addition, the experiments switch to a size-agnostic setting, making the 1.0% and 3.2% improvements difficult to compare with the size-aware RANK-based evaluation in §4.1.4. Without a clean held-out protocol and variance estimates, the claim that the synthesized queue topologies outperform all seventeen baselines is not supported.
  5. [§5.2] The memory-tiering evaluation compares only against "vanilla ARMS" on four workloads and reports improvements of 2.5–7.9%. No comparison is made to Memtis or other state-of-the-art tiering policies, no repeated runs or confidence intervals are reported, and there is no ablation separating the effect of the synthesized policy from the effect of the richer 20-window access history added by VULCAN. These small gains need considerably more experimental support before the paper can claim superiority over existing tiering systems.
minor comments (4)
  1. [§3.2.1] The runtime instance classifier shown in Figure 5 is described but never evaluated. If the paper claims to support runtime instance detection and policy selection, this component needs at least a proof-of-concept measurement.
  2. [§4.1.2] The choice of K=10 for KMeans and the selection of fifteen trace features are not validated. The paper should justify the cluster count and feature set, and ideally show sensitivity to these choices, especially since clusters define the notion of "instance." The use of only the first 50,000 requests per trace (<1% of the trace) also deserves a representativeness check.
  3. [Table 4] Table 4 lists congestion control as a possible instantiation with an eBPF-based policy module and Mahimahi evaluation, but no congestion-control experiment appears in the paper. Either remove the row or add the corresponding evaluation.
  4. [Appendix B, Listing 3] The prompt text contains a duplicated line describing the history metadata ("auto info = history.get_metadata(obj_id)" appears twice), and the prose in the same appendix has a small typo ("some some illustrative examples"). These should be cleaned up.

Circularity Check

1 steps flagged · score 6.0 of 10

Cache evaluation is polluted by the search trace: cluster-average MRR includes the trace used to fit each heuristic, so the 1.94–69% gains are partly a restatement of the optimization objective.

  1. fitted input called prediction [§4.1.3–§4.1.4 (cache eviction policy search and results)]
    "During the search phase, the evaluator harness uses one trace from within the cluster to score candidate solutions – the heuristic identified at the end of the search is then evaluated on all traces within this cluster."

    The paper's headline evidence (§4.1.4) is cluster-average MRR over 'all traces in a cluster,' and the heuristic was selected by maximizing hit rate/MRR on one undisclosed trace inside that same cluster. The search objective is therefore included in the reported average, so part of the claimed 1.94–69% improvement over baselines is mechanically the fitness function used to choose the heuristic. No held-out trace split or disclosure of the search trace's identity is provided, so the instance-generalization claim is not independently demonstrated.

full rationale

The only circularity I can exhibit by construction is the cache-evaluation protocol. The rest of the derivation chain — VALUE/RANK interfaces, policy/mechanism separation, and evolutionary search over LLM-generated scoring functions — does not reduce to its inputs. The self-citations ([27], [46], [47], [88], [109]) are code availability, motivation, related-work framing, or an experimentally measured ARMS baseline; none is used as a uniqueness theorem or to forbid alternatives, so none is load-bearing circularity. Separate reporting problems exist but are not circularity: the abstract's Anvil safety guarantee and 4.9x spot-VM claim are absent from the body, and the abstract's cache/tiering numbers (2x, 10%) do not match the body's 1.94–69% and 2.5–7.9%. The queue-topology results (§4.2) reuse the same cluster definitions and per-instance evaluation, so they inherit the same train/test caveat when the same one-trace scorer is used. Overall, one central quantitative claim partially reduces to its optimization objective, warranting 6 rather than 0–2.

Assumptions & free parameters 6 free parameters · 6 assumptions · 0 invented entities

The paper's contribution is mostly an interface/framing; the reported numbers are produced by evolutionary search against the same instances on which they are evaluated. The safety guarantee is assumed from the interface shape rather than proven, instance boundaries rest on unvalidated clustering, and the discovered heuristics' constants are free parameters optimized on the training trace.

free parameters (6)
  • KMeans cluster count K = 10
    Defines the 10 cache instances; no sensitivity analysis is given, and the number of clusters directly shapes the per-cluster results.
  • Hand-selected 15 trace features for clustering = not enumerated
    Feature choice for KMeans is ad hoc and never listed; changing these features changes instance boundaries and therefore reported gains.
  • Evolutionary search hyperparameters = 25 candidates/round; top-2 retained; 150 iterations for tiering; cache round count undisclosed
    Search settings are chosen without justification. The cache search round count is never stated, so the search budget is not reproducible.
  • Single training trace per cluster = one trace per cluster (identity undisclosed)
    The heuristic is fitted on this trace (§4.1.3); its identity is not disclosed, so cluster-average results cannot be separated from overfitting.
  • Constants in discovered heuristics = e.g., Q0_PROMOTE_THRESHOLD=2, Q1_PROMOTE_THRESHOLD=1, Q2_STALE_AGE=100000; nvm_bw_penalty=0.55/0.8/0.92; phase_penalty=0
    These constants are direct outputs of the search on the training instance — fitted parameters, not predictions. Listing them as discovered behavior does not make them independent.
  • Memory-tiering per-page window length = 20 windows (10 s)
    Vulcan modifies ARMS data collection to a 20-window circular buffer (§5.1.1); this choice is not derived or compared against other buffer sizes.
assumptions (6)
  • domain assumption A stateless function that returns a numeric value cannot be an invalid policy; safety reduces to well-typedness.
    §3.1 states 'every function that returns a real value is a well-formed policy... it cannot be an invalid one.' This assumes semantic safety (no crashes, no invariant violations) follows from the interface shape; no proof is provided.
  • ad hoc to paper KMeans clusters of CloudPhysics traces correspond to meaningful deployment instances for which one specialized heuristic is appropriate.
    §4.1.2 clusters traces using KMeans on 15 features from the first 50k requests; cluster coherence, stability, and representativeness are not validated, despite 'instance' being the central unit of the 'instance-optimal' claim.
  • domain assumption Object hit rate / miss-rate reduction over FIFO is the right objective for cache performance.
    §4.1.4 and §4.2.4 use MRR; §4.2 additionally ignores object sizes to match slab-based literature. No end-to-end latency, bandwidth cost, or user-facing metric is used, so 'performance' is a proxy.
  • domain assumption ARMS's access tracking and migration mechanism is sound; Vulcan need only replace the scoring policy.
    §5.1.1: 'We replace the policy components with a single policy module, while leaving the action component untouched.' If ARMS's mechanism has flaws, Vulcan inherits them; no independent validation of the mechanism is reported.
  • ad hoc to paper The queue-topology search space (M≤5 FIFO/LRU queues plus one ghost queue) contains performant, efficient-by-design eviction policies.
    §4.2.1 defines this space from the observation that many algorithms are queue-based; there is no completeness or optimality argument for the restriction M≤5, ghost capacity ≤L, or the specific transition functions.
  • domain assumption The LLM-assisted survey classification of 660 OSDI/NSDI papers is accurate.
    Appendix A uses an LLM pipeline with no human agreement check or error analysis; the counts (191 papers, 234 tasks, 71 VALUE / 158 RANK) are presented without validation and underpin the paper's generality claim.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Vulcan: Instance-specialized, Verifiable Systems Heuristics Through LLM-driven Search." pith.science (2026). https://pith.science/paper/Z5TAER47

@misc{pith2026251225065,
  author       = {Pith},
  title        = {Pith review of: Vulcan: Instance-specialized, Verifiable Systems Heuristics Through LLM-driven Search},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Z5TAER47}},
  note         = {Machine review of arXiv:2512.25065}
}
read the original abstract

Systems resource management tasks rely primarily on hand-designed heuristics. However, growing hardware heterogeneity and workload diversity require heuristics specialized to particular deployment instances, making manual design expensive and difficult to scale. In this paper, we explore how to synthesize systems heuristics using LLMs. The main challenge is ensuring that generated heuristics execute safely, integrate correctly with the surrounding system, and still achieve strong performance. We propose Vulcan, a framework that identifies LLM-friendly interfaces that isolate core decision logic from the rest of the implementation. With Vulcan, LLM-generated code is restricted to simple stateless decision functions, while trusted runtime abstractions provide rich derived statistics for meaningful policy exploration without system-integration bugs. To ensure execution safety, LLMs synthesize heuristics in a restricted language, Anvil, that guarantees important properties by construction. We evaluate Vulcan across three well-studied domains and demonstrate up to 4.9x higher savings for spot-VM scheduling, up to 2x lower miss ratios for cache eviction, and up to 10% higher application performance for tiered-memory systems, while ensuring execution safety throughout.

Figures

Figures reproduced from arXiv: 2512.25065 by the authors.

Figure 1
Figure 1. Count of CloudPhysics traces where each heuristic per￾forms best (highest object hit rate). Tiny, small, and large caches correspond to 0.1%, 1%, and 10% of the trace footprint (i.e., num￾ber of unique objects in a trace); we ignore object sizes, similar to how [120] and [115] report their results. The column “Others” in￾cludes SR_LRU [85], CR_LFU [85], Sieve [120], and GDSF [23] all of which had at least one trace … view at source ↗
Figure 2
Figure 2. Three-step user-facing pipeline for instantiating VUL￾CAN. All of these are part of the problem specification phase (Phase I) shown in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. End-to-end overview of VULCAN. Each of the inputs defined in Phase I are used by subsequent phases [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Division of labour in VULCAN: users are responsible for implementing mechanisms for data collection and action (in green); LLMs synthesize the scoring functions (in yellow); VULCAN uses templates to assemble these LLM-generated functions into an executable policy modul…
Figure 5
Figure 5. Figure 5: Automated instance generation and runtime policy selec [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Search process in VULCAN. Green boxes (template, evalu￾ation harness) are specified by users of VULCAN; blue boxes (gener￾ator, heuristic DB, module injection) are implemented by VULCAN. additional constraints in natural language that the generated function must respec…
Figure 7
Figure 7. Figure 7: t-SNE visualization of the CloudPhysics dataset, with points colored by cluster assignment (0 – 9). covered through the search, is provided in Appendix B. For our evaluator harness, we opted to use a simulator – lib￾cachesim [45] – which runs the newly generated heuris…
Figure 8
Figure 8. Figure 8: Performance of VULCAN-synthesized cache eviction heuristics compared against baselines, per cluster (C0 – C9). Higher is better. We instantiated a VULCAN search on each of the clusters (i.e., instances) independently, and discovered a specialized heuristic for each of …
Figure 9
Figure 9. Figure 9: Two cache queue topologies discovered by [PITH_FULL_IMAGE:figures/full_fig_p011_9.png]
Figure 10
Figure 10. Figure 10: Performance of VULCAN-synthesized Queue Topology heuristics compared against queue based baselines on clusters C7 and C8. In both clusters C7 and C8, the heuristic synthesized by VULCAN outperforms all seventeen baselines considered. Specifically, VULCAN improves over…
Figure 11
Figure 11. Figure 11: ) – driving the cost of discovery down further. Deep-dive into the generated heuristics. We find that VUL￾CAN can discover interesting ways of fine-tuning heuristics. Often times, these heuristics exceed 150 lines of code – show￾casing that VULCAN’s evolutionary searc…

Discussion (0). Sign in to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Defining AI-Native Systems: Autonomy as Revision Authority

    cs.AI 2026-07 accept novelty 7.0 of 10

    AI-native systems are defined by an AI holding autonomous revision authority over the system's own implementation, verified by a fallback and escalation detector, with human ownership of purpose.

Reference graph

Works this paper leans on

129 extracted references · 14 linked inside Pith · cited by 1 Pith paper

  1. [27]

    Man-made heuristics are dead

    Rohit Dwivedula, Divyanshu Saxena, Aditya Akella, Swarat Chaudhuri, and Daehyeok Kim. Man-made heuristics are dead. long live code generators! InPro- ceedings of the 24th ACM Workshop on Hot Topics in Networks, pages 51–60, 2025

    2025

  2. [90]

    Openevolve: An open-source evolutionary coding agent

    Asankhaya Sharma. Openevolve: An open-source evolutionary coding agent. https://github.com/ codelion/openevolve, 2025. Accessed: 2025-12-10

    2025

  3. [1]

    https://kubernetes.io/ docs/tasks/run-application/horizontal-pod- autoscale/

    Horizontal pod autoscale. https://kubernetes.io/ docs/tasks/run-application/horizontal-pod- autoscale/. Accessed: Dec 2025

    2025

  4. [2]

    C2tcp: A flexible cellular tcp to meet stringent delay requirements.IEEE Journal on Selected Areas in Com- munications, 37(4):918–932, 2019

    Soheil Abbasloo, Yang Xu, and H Jonathan Chao. C2tcp: A flexible cellular tcp to meet stringent delay requirements.IEEE Journal on Selected Areas in Com- munications, 37(4):918–932, 2019

    2019

  5. [3]

    Jonathan Chao

    Soheil Abbasloo, Chen-Yu Yen, and H. Jonathan Chao. Classic meets modern: a pragmatic learning-based con- gestion control for the internet. InProceedings of the Annual Conference of the ACM Special Interest Group on Data Communication on the Applications, Tech- nologies, Architectures, and Protocols for Computer Communication, SIGCOMM ’20, page 632–647, N...

    2020

  6. [4]

    Mitosis work- load btree, 2019

    Reto Achermann and Ashish Panwar. Mitosis work- load btree, 2019

    2019

  7. [5]

    Vidur: A large-scale simulation framework for llm inference

    Amey Agrawal, Nitin Kedia, Jayashree Mohan, Ashish Panwar, Nipun Kwatra, Bhargav S Gulavani, Ra- machandran Ramjee, and Alexey Tumanov. Vidur: A large-scale simulation framework for llm inference. Proceedings of Machine Learning and Systems, 6:351– 366, 2024

    2024

  8. [6]

    Kmlib: Towards machine learning for oper- ating systems

    Ibrahim Umit Akgun, Ali Selman Aydin, and Erez Zadok. Kmlib: Towards machine learning for oper- ating systems. InProceedings of the On-Device Intel- ligence Workshop, co-located with the MLSys Confer- ence, pages 1–6, 2020

    2020

Show all 129 references
  1. [7]

    CherryPick: Adaptively unearthing the best cloud configurations for big data analytics

    Omid Alipourfard, Hongqiang Harry Liu, Jianshu Chen, Shivaram Venkataraman, Minlan Yu, and Ming Zhang. CherryPick: Adaptively unearthing the best cloud configurations for big data analytics. In 14th USENIX Symposium on Networked Systems De- sign and Implementation (NSDI 17), p...

    2017

  2. [8]

    Maltz, Jitendra Padhye, Parveen Patel, Balaji Prab- hakar, Sudipta Sengupta, and Murari Sridharan

    Mohammad Alizadeh, Albert Greenberg, David A. Maltz, Jitendra Padhye, Parveen Patel, Balaji Prab- hakar, Sudipta Sengupta, and Murari Sridharan. Data center tcp (dctcp). InProceedings of the ACM SIG- COMM 2010 Conference, SIGCOMM ’10, page 63–74, New York, NY , USA, 2010. Asso...

    2010

  3. [9]

    Starvation in end-to-end congestion control

    Venkat Arun, Mohammad Alizadeh, and Hari Balakr- ishnan. Starvation in end-to-end congestion control. In Proceedings of the ACM SIGCOMM 2022 Conference, SIGCOMM ’22, page 177–192, New York, NY , USA,

    2022

  4. [10]

    Copa: Practical {Delay-Based} congestion control for the internet

    Venkat Arun and Hari Balakrishnan. Copa: Practical {Delay-Based} congestion control for the internet. In 15th USENIX Symposium on Networked Systems De- sign and Implementation (NSDI 18), pages 329–342, Renton, W A, 2018. USENIX Association

    2018

  5. [11]

    Caching with delayed hits

    Nirav Atre, Justine Sherry, Weina Wang, and Daniel S Berger. Caching with delayed hits. InProceedings of the Annual conference of the ACM Special Interest Group on Data Communication on the applications, technologies, architectures, and protocols for computer communication, pa...

    2020

  6. [12]

    Pro- gram synthesis with large language models.arXiv preprint arXiv:2108.07732, 2021

    Jacob Austin, Augustus Odena, Maxwell Nye, Maarten Bosma, Henryk Michalewski, David Dohan, Ellen Jiang, Carrie Cai, Michael Terry, Quoc Le, et al. Pro- gram synthesis with large language models.arXiv preprint arXiv:2108.07732, 2021

    2021 arXiv

  7. [13]

    {LHD}: Improving cache hit rate by maximizing hit density

    Nathan Beckmann, Haoxian Chen, and Asaf Cidon. {LHD}: Improving cache hit rate by maximizing hit density. In15th USENIX Symposium on Networked Systems Design and Implementation (NSDI 18), pages 389–403, Renton, W A, 2018. USENIX Association

    2018

  8. [14]

    Laszlo A. Belady. A study of replacement algorithms for a virtual-storage computer.IBM Systems journal, 5(2):78–101, 1966

    1966

  9. [15]

    {RobinHood}: Tail latency aware caching–dynamic reallocation from {Cache-Rich} to {Cache-Poor}

    Daniel S Berger, Benjamin Berg, Timothy Zhu, Sid- dhartha Sen, and Mor Harchol-Balter. {RobinHood}: Tail latency aware caching–dynamic reallocation from {Cache-Rich} to {Cache-Poor}. In13th USENIX Sym- posium on Operating Systems Design and Implementa- tion (OSDI 18), pages 19...

    2018

  10. [16]

    Hyperbolic caching: Flexible caching for web applications

    Aaron Blankstein, Siddhartha Sen, and Michael J Freedman. Hyperbolic caching: Flexible caching for web applications. In2017 USENIX Annual Technical Conference (USENIX ATC 17), pages 499–511, 2017

    2017

  11. [17]

    FetchBPF: Customiz- able prefetching policies in linux with eBPF

    Xuechun Cao, Shaurya Patel, Soo Yee Lim, Xueyuan Han, and Thomas Pasquier. FetchBPF: Customiz- able prefetching policies in linux with eBPF. In 2024 USENIX Annual Technical Conference (USENIX ATC 24), pages 369–378, Santa Clara, CA, July 2024. USENIX Association

    2024

  12. [18]

    Bbr: Congestion-based congestion control.Communica- tions of the ACM, 60(2):58–66, 2017

    Neal Cardwell, Yuchung Cheng, C Stephen Gunn, Soheil Hassas Yeganeh, and Van Jacobson. Bbr: Congestion-based congestion control.Communica- tions of the ACM, 60(2):58–66, 2017

    2017

  13. [19]

    Evo- prompting: Language models for code-level neural ar- chitecture search.Advances in neural information processing systems, 36:7787–7817, 2023

    Angelica Chen, David Dohan, and David So. Evo- prompting: Language models for code-level neural ar- chitecture search.Advances in neural information processing systems, 36:7787–7817, 2023

    2023

  14. [20]

    Darwin: Flexible learning- based cdn caching

    Jiayi Chen, Nihal Sharma, Tarannum Khan, Shu Liu, Brian Chang, Aditya Akella, Sanjay Shakkottai, and Ramesh K Sitaraman. Darwin: Flexible learning- based cdn caching. InProceedings of the ACM SIG- COMM 2023 Conference, ACM SIGCOMM ’23, page 981–999, New York, NY , USA, 2023. A...

    2023

  15. [21]

    Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brock- man, Alex Ray, Raul Puri, Gretchen Krueger, Michael Petrov, Heidy Khlaaf, Girish Sastry, Pamela Mishkin, Brooke Chan, Scott...

    2021

  16. [22]

    Barbarians at the gate: How ai is upending systems research.arXiv preprint arXiv:2510.06189, 2025

    Audrey Cheng, Shu Liu, Melissa Pan, Zhifei Li, Bowen Wang, Alex Krentsel, Tian Xia, Mert Cemri, Jongseok Park, Shuo Yang, et al. Barbarians at the gate: How ai is upending systems research.arXiv preprint arXiv:2510.06189, 2025

    2025

  17. [23]

    Hewlett-Packard Laboratories, Palo Alto, CA, USA, 1998

    Ludmila Cherkasova.Improving WWW proxies perfor- mance with greedy-dual-size-frequency caching policy. Hewlett-Packard Laboratories, Palo Alto, CA, USA, 1998

    1998

  18. [24]

    Massachusetts Institute of Technology, Cambridge, MA, 1968

    Fernando J Corbato.A paging experiment with the multics system. Massachusetts Institute of Technology, Cambridge, MA, 1968

    1968

  19. [25]

    Let them run cake, Jun 2018

    Jonathan Corbet. Let them run cake, Jun 2018

    2018

  20. [26]

    The design and operation of {CloudLab}

    Dmitry Duplyakin, Robert Ricci, Aleksander Maricq, Gary Wong, Jonathon Duerig, Eric Eide, Leigh Stoller, Mike Hibler, David Johnson, Kirk Webb, et al. The design and operation of {CloudLab}. In2019 USENIX annual technical conference (USENIX ATC 19), pages 1–14, 2019

    2019

  21. [28]

    Tinylfu: A highly efficient cache admission policy.ACM Trans- actions on Storage (ToS), 13(4):1–31, 2017

    Gil Einziger, Roy Friedman, and Ben Manes. Tinylfu: A highly efficient cache admission policy.ACM Trans- actions on Storage (ToS), 13(4):1–31, 2017

    2017

  22. [29]

    Rossbach

    Henrique Fingler, Isha Tarte, Hangchen Yu, Ariel Szekely, Bodun Hu, Aditya Akella, and Christopher J. Rossbach. Towards a machine learning-assisted kernel with lake. InProceedings of the 28th ACM Interna- tional Conference on Architectural Support for Pro- gramming Languages a...

    2023

  23. [30]

    The championship simulator: Architec- tural simulation for education and competition.arXiv preprint arXiv:2210.14324, 2022

    Nathan Gober, Gino Chacon, Lei Wang, Paul V Gratz, Daniel A Jimenez, Elvira Teran, Seth Pugsley, and Jinchun Kim. The championship simulator: Architec- tural simulation for education and competition.arXiv preprint arXiv:2210.14324, 2022

    2022 arXiv

  24. [31]

    Goyal, H.M

    P. Goyal, H.M. Vin, and Haichen Cheng. Start-time fair queueing: a scheduling algorithm for integrated services packet switching networks.IEEE/ACM Trans- actions on Networking, 5(5):690–704, 1997

    1997

  25. [32]

    Altruistic scheduling in Multi-Resource clusters

    Robert Grandl, Mosharaf Chowdhury, Aditya Akella, and Ganesh Ananthanarayanan. Altruistic scheduling in Multi-Resource clusters. In12th USENIX Sympo- sium on Operating Systems Design and Implementa- tion (OSDI 16), pages 65–80, Savannah, GA, Novem- ber 2016. USENIX Association. 16

    2016

  26. [33]

    GRAPHENE: Pack- ing and Dependency-Aware scheduling for Data- Parallel clusters

    Robert Grandl, Srikanth Kandula, Sriram Rao, Aditya Akella, and Janardhan Kulkarni. GRAPHENE: Pack- ing and Dependency-Aware scheduling for Data- Parallel clusters. In12th USENIX Symposium on Op- erating Systems Design and Implementation (OSDI 16), pages 81–97, Savannah, GA, N...

    2016

  27. [34]

    Cubic: a new tcp-friendly high-speed tcp variant.ACM SIGOPS operating systems review, 42(5):64–74, 2008

    Sangtae Ha, Injong Rhee, and Lisong Xu. Cubic: a new tcp-friendly high-speed tcp variant.ACM SIGOPS operating systems review, 42(5):64–74, 2008

    2008

  28. [35]

    Glia: A human-inspired ai for automated systems design and optimization.arXiv preprint arXiv:2510.27176, 2025

    Pouya Hamadanian, Pantea Karimi, Arash Nasr- Esfahany, Kimia Noorbakhsh, Joseph Chandler, Ali ParandehGheibi, Mohammad Alizadeh, and Hari Bal- akrishnan. Glia: A human-inspired ai for automated systems design and optimization.arXiv preprint arXiv:2510.27176, 2025

    2025 arXiv

  29. [36]

    Congestion control system opti- mization with large language models.arXiv preprint arXiv:2508.16074, 2025

    Zhiyuan He, Aashish Gottipati, Lili Qiu, Yuqing Yang, and Francis Y Yan. Congestion control system opti- mization with large language models.arXiv preprint arXiv:2508.16074, 2025

    2025 arXiv

  30. [37]

    Mea- suring coding challenge competence with apps.arXiv preprint arXiv:2105.09938, 2021

    Dan Hendrycks, Steven Basart, Saurav Kadavath, Man- tas Mazeika, Akul Arora, Ethan Guo, Collin Burns, Samir Puranik, Horace He, Dawn Song, et al. Mea- suring coding challenge competence with apps.arXiv preprint arXiv:2105.09938, 2021

    2021 arXiv

  31. [38]

    An analysis of facebook photo caching

    Qi Huang, Ken Birman, Robbert Van Renesse, Wyatt Lloyd, Sanjeev Kumar, and Harry C Li. An analysis of facebook photo caching. InProceedings of the Twenty-Fourth ACM Symposium on Operating Systems Principles, pages 167–181, 2013

    2013

  32. [39]

    Intel 64 and ia-32 architectures software developer manuals

    Intel Corporation. Intel 64 and ia-32 architectures software developer manuals. 2018

    2018

  33. [40]

    Amazon nova premier: Technical report and model card

    Amazon Artificial General Intelligence. Amazon nova premier: Technical report and model card. 2025

    2025

  34. [41]

    Live- codebench: Holistic and contamination free evaluation of large language models for code.arXiv preprint arXiv:2403.07974, 2024

    Naman Jain, King Han, Alex Gu, Wen-Ding Li, Fanjia Yan, Tianjun Zhang, Sida Wang, Armando Solar-Lezama, Koushik Sen, and Ion Stoica. Live- codebench: Holistic and contamination free evaluation of large language models for code.arXiv preprint arXiv:2403.07974, 2024

    2024 arXiv

  35. [42]

    Clock- pro: An effective improvement of the clock replace- ment

    Song Jiang, Feng Chen, and Xiaodong Zhang. Clock- pro: An effective improvement of the clock replace- ment. InUSENIX Annual Technical Conference, Gen- eral Track, pages 323–336, 2005

    2005

  36. [43]

    Lirs: An efficient low inter-reference recency set replacement policy to improve buffer cache performance.ACM SIGMET- RICS Performance Evaluation Review, 30(1):31–42, 2002

    Song Jiang and Xiaodong Zhang. Lirs: An efficient low inter-reference recency set replacement policy to improve buffer cache performance.ACM SIGMET- RICS Performance Evaluation Review, 30(1):31–42, 2002

    2002

  37. [44]

    2q: A low over- head high performance buffer management replace- ment algorithm

    Theodore Johnson and Dennis Shasha. 2q: A low over- head high performance buffer management replace- ment algorithm. InProceedings of the 20th Interna- tional Conference on Very Large Data Bases, VLDB ’94, page 439–450, San Francisco, CA, USA, 1994. Morgan Kaufmann Publishers Inc

    1994

  38. [45]

    libcachesim: a high perfor- mance library for building cache simulators

    Juncheng Yang (1a1a11a). libcachesim: a high perfor- mance library for building cache simulators. https: //github.com/1a1a11a/libCacheSim, 2023. Accessed: 2025-06-28

    2023

  39. [46]

    Herding lla- mas: Using llms as an os module.arXiv preprint arXiv:2401.08908, 2024

    Aditya K Kamath and Sujay Yadalam. Herding lla- mas: Using llms as an os module.arXiv preprint arXiv:2401.08908, 2024

    2024 arXiv

  40. [47]

    Striking the right chord: Parameter tuning in memory tiering systems

    Konstantinos Kanellis, Sujay Yadalam, Shivaram Venkataraman, and Michael Swift. Striking the right chord: Parameter tuning in memory tiering systems. In Proceedings of the 3rd Workshop on Disruptive Mem- ory Systems, pages 1–9, 2025

    2025

  41. [48]

    Caching strategies to improve disk system performance.Computer, 27(3):38–46, 1994

    Ramakrishna Karedla, J Spencer Love, and Bradley G Wherry. Caching strategies to improve disk system performance.Computer, 27(3):38–46, 1994

    1994

  42. [49]

    Robust heuristic algorithm design with llms

    Pantea Karimi, Dany Rouhana, Pooria Namyar, Siva Kesava Reddy Kakarla, Venkat Arun, and Behnaz Arzani. Robust heuristic algorithm design with llms. arXiv preprint arXiv:2510.08755, 2025

    2025

  43. [50]

    Crust- bench: A comprehensive benchmark for c-to-safe-rust transpilation.arXiv preprint arXiv:2504.15254, 2025

    Anirudh Khatry, Robert Zhang, Jia Pan, Ziteng Wang, Qiaochu Chen, Greg Durrett, and Isil Dillig. Crust- bench: A comprehensive benchmark for c-to-safe-rust transpilation.arXiv preprint arXiv:2504.15254, 2025

    2025

  44. [51]

    Gautam Kumar, Nandita Dukkipati, Keon Jang, Hassan M. G. Wassel, Xian Wu, Behnam Montazeri, Yaogong Wang, Kevin Springborn, Christopher Alfeld, Michael Ryan, David Wetherall, and Amin Vahdat. Swift: De- lay is simple and effective for congestion control in the datacenter. InPr...

    2020

  45. [52]

    Robus: fair cache allocation for data-parallel workloads

    Mayuresh Kunjir, Brandon Fain, Kamesh Munagala, and Shivnath Babu. Robus: fair cache allocation for data-parallel workloads. InProceedings of the 2017 ACM International Conference on Management of Data, pages 219–234, New York, NY , USA, 2017. As- sociation for Computing Machinery. 17

    2017

  46. [53]

    Kurniawan, Rani Ayu Putri, Peiran Qin, Kahfi S

    Daniar H. Kurniawan, Rani Ayu Putri, Peiran Qin, Kahfi S. Zulkifli, Ray A. O. Sinurat, Janki Bhimani, Sandeep Madireddy, Achmad Imam Kistijantoro, and Haryadi S. Gunawi. Heimdall: Optimizing storage i/o admission with extensive machine learning pipeline. InProceedings of the T...

    2025

  47. [54]

    Shinkaevolve: Towards open-ended and sample-efficient program evolution.arXiv preprint arXiv:2509.19349, 2025

    Robert Tjarko Lange, Yuki Imajuku, and Edoardo Cetin. Shinkaevolve: Towards open-ended and sample-efficient program evolution.arXiv preprint arXiv:2509.19349, 2025

    2025 arXiv

  48. [55]

    Memtis: Efficient memory tiering with dynamic page classification and page size deter- mination

    Taehyung Lee, Sumit Kumar Monga, Changwoo Min, and Young Ik Eom. Memtis: Efficient memory tiering with dynamic page classification and page size deter- mination. InProceedings of the 29th Symposium on Operating Systems Principles, 2023

    2023

  49. [56]

    Bush, Prakash Ramanan, Rajesh Kumar, Thomas Chestna, Yajing Liu, Ying Liu, Ye Zhao, Kathryn S

    Jianheng Ling, Pratik Worah, Yawen Wang, Yunchuan Kong, Anshul Kapoor, Chunlei Wang, Clifford Stein, Diwakar Gupta, Jason Behmer, Logan A. Bush, Prakash Ramanan, Rajesh Kumar, Thomas Chestna, Yajing Liu, Ying Liu, Ye Zhao, Kathryn S. McKinley, Meeyoung Park, and Martin Maas. L...

    2025

  50. [57]

    The linux scheduler: a decade of wasted cores

    Jean-Pierre Lozi, Baptiste Lepers, Justin Funston, Fa- bien Gaud, Vivien Quéma, and Alexandra Fedorova. The linux scheduler: a decade of wasted cores. In Proceedings of the Eleventh European Conference on Computer Systems, pages 1–16, 2016

    2016

  51. [58]

    Visu- alizing data using t-sne.Journal of machine learning research, 9(Nov):2579–2605, 2008

    Laurens van der Maaten and Geoffrey Hinton. Visu- alizing data using t-sne.Journal of machine learning research, 9(Nov):2579–2605, 2008

    2008

  52. [59]

    Mark Mansi, Bijan Tabatabai, and Michael M. Swift. CBMM: Financial advice for kernel memory man- agers. In2022 USENIX Annual Technical Conference (USENIX ATC 22), pages 593–608, Carlsbad, CA, July

  53. [60]

    Learning scheduling algorithms for data processing clusters

    Hongzi Mao, Malte Schwarzkopf, Shaileshh Bojja Venkatakrishnan, Zili Meng, and Mohammad Alizadeh. Learning scheduling algorithms for data processing clusters. InProceedings of the ACM Special Interest Group on Data Communication, SIGCOMM ’19, page 270–288, New York, NY , USA, ...

    2019

  54. [61]

    Tpp: Transparent page place- ment for cxl-enabled tiered-memory

    Hasan Al Maruf, Hao Wang, Abhishek Dhanotia, Jo- hannes Weiner, Niket Agarwal, Pallab Bhattacharya, Chris Petersen, Mosharaf Chowdhury, Shobhit Kanau- jia, and Prakash Chauhan. Tpp: Transparent page place- ment for cxl-enabled tiered-memory. InProceedings of the 28th ACM Inter...

    2023

  55. [62]

    {ARC}: A {Self-Tuning}, low overhead replacement cache

    Nimrod Megiddo and Dharmendra S Modha. {ARC}: A {Self-Tuning}, low overhead replacement cache. In 2nd USENIX Conference on File and Storage Tech- nologies (FAST 03), San Francisco, CA, USA, 2003. USENIX Association

    2003

  56. [63]

    Interpreting deep learning-based networking systems

    Zili Meng, Minhu Wang, Jiasong Bai, Mingwei Xu, Hongzi Mao, and Hongxin Hu. Interpreting deep learning-based networking systems. InProceedings of the Annual Conference of the ACM Special Interest Group on Data Communication on the Applications, Technologies, Architectures, and...

    2020

  57. [64]

    Best-offset hardware prefetching

    Pierre Michaud. Best-offset hardware prefetching. In 2016 IEEE International Symposium on High Perfor- mance Computer Architecture (HPCA), pages 469–480. IEEE, 2016

    2016

  58. [65]

    Timely: Rtt-based congestion control for the dat- acenter

    Radhika Mittal, Vinh The Lam, Nandita Dukkipati, Emily Blem, Hassan Wassel, Monia Ghobadi, Amin Vahdat, Yaogong Wang, David Wetherall, and David Zats. Timely: Rtt-based congestion control for the dat- acenter. InProceedings of the 2015 ACM Conference on Special Interest Group ...

    2015

  59. [66]

    Towards automated verification of llm-synthesized c programs, 2025

    Prasita Mukherjee and Benjamin Delaware. Towards automated verification of llm-synthesized c programs, 2025

    2025

  60. [67]

    Heterogeneity-Aware cluster scheduling policies for deep learning workloads

    Deepak Narayanan, Keshav Santhanam, Fiodar Kazhamiaka, Amar Phanishayee, and Matei Zaharia. Heterogeneity-Aware cluster scheduling policies for deep learning workloads. In14th USENIX Symposium on Operating Systems Design and Implementation (OSDI 20), pages 481–498. USENIX Asso...

    2020

  61. [68]

    Write off-loading: Practical power man- agement for enterprise storage.ACM Transactions on Storage (TOS), 4(3):1–23, 2008

    Dushyanth Narayanan, Austin Donnelly, and Antony Rowstron. Write off-loading: Practical power man- agement for enterprise storage.ACM Transactions on Storage (TOS), 4(3):1–23, 2008

    2008

  62. [69]

    Usama Naseer and Theophilus A. Benson. Configana- tor: A data-driven approach to improving CDN per- formance. In19th USENIX Symposium on Networked Systems Design and Implementation (NSDI 22), pages 1135–1158, Renton, W A, April 2022. USENIX Asso- ciation. 18

    2022

  63. [70]

    Mahimahi: accurate {Record-and- Replay} for {HTTP}

    Ravi Netravali, Anirudh Sivaraman, Somak Das, Ameesh Goyal, Keith Winstein, James Mickens, and Hari Balakrishnan. Mahimahi: accurate {Record-and- Replay} for {HTTP}. In2015 USENIX Annual Tech- nical Conference (USENIX ATC 15), pages 417–429, Santa Clara, CA, USA, 2015. USENIX ...

    2015

  64. [71]

    Alexander Novikov, Ngân V ˜u, Marvin Eisenberger, Emilien Dupont, Po-Sen Huang, Adam Zsolt Wag- ner, Sergey Shirobokov, Borislav Kozlovskii, Fran- cisco J. R. Ruiz, Abbas Mehrabian, M. Pawan Ku- mar, Abigail See, Swarat Chaudhuri, George Holland, Alex Davies, Sebastian Nowozin...

    2025

  65. [72]

    The akamai network: a platform for high-performance internet applications.ACM SIGOPS Operating Sys- tems Review, 44(3):2–19, 2010

    Erik Nygren, Ramesh K Sitaraman, and Jennifer Sun. The akamai network: a platform for high-performance internet applications.ACM SIGOPS Operating Sys- tems Review, 44(3):2–19, 2010

    2010

  66. [73]

    O’Neil, Patrick E

    Elizabeth J. O’Neil, Patrick E. O’Neil, and Ger- hard Weikum. The lru-k page replacement algo- rithm for database disk buffering.SIGMOD Rec., 22(2):297–306, June 1993

    1993

  67. [74]

    Sparrow: distributed, low latency schedul- ing

    Kay Ousterhout, Patrick Wendell, Matei Zaharia, and Ion Stoica. Sparrow: distributed, low latency schedul- ing. InProceedings of the twenty-fourth ACM sym- posium on operating systems principles, pages 69–84, 2013

    2013

  68. [75]

    Kernelbench: Can llms write efficient gpu ker- nels?arXiv preprint arXiv:2502.10517, 2025

    Anne Ouyang, Simon Guo, Simran Arora, Alex L Zhang, William Hu, Christopher Ré, and Azalia Mirho- seini. Kernelbench: Can llms write efficient gpu ker- nels?arXiv preprint arXiv:2502.10517, 2025

    2025 arXiv

  69. [76]

    Completely fair scheduler

    Chandandeep Singh Pabla. Completely fair scheduler. Linux Journal, 2009(184):4, 2009

    2009

  70. [77]

    Mutant: Learning congestion control from exist- ing protocols via online reinforcement learning

    Lorenzo Pappone, Alessio Sacco, and Flavio Espos- ito. Mutant: Learning congestion control from exist- ing protocols via online reinforcement learning. In 22nd USENIX Symposium on Networked Systems De- sign and Implementation (NSDI 25), pages 1507–1522, 2025

    2025

  71. [78]

    Pro- filing dynamic data access patterns with controlled overhead and quality

    SeongJae Park, Yunjae Lee, and Heon Y Yeom. Pro- filing dynamic data access patterns with controlled overhead and quality. InProceedings of the 20th In- ternational Middleware Conference Industrial Track, 2019

    2019

  72. [79]

    Real-time dy- namic voltage scaling for low-power embedded oper- ating systems

    Padmanabhan Pillai and Kang G Shin. Real-time dy- namic voltage scaling for low-power embedded oper- ating systems. InProceedings of the eighteenth ACM symposium on Operating systems principles, pages 89– 102, 2001

    2001

  73. [80]

    A sim- ple synchronous distributed-memory algorithm for the hpcc randomaccess benchmark

    Steven J Plimpton, Ron Brightwell, Courtenay Vaughan, Keith Underwood, and Mike Davis. A sim- ple synchronous distributed-memory algorithm for the hpcc randomaccess benchmark. In2006 IEEE Inter- national Conference on Cluster Computing, pages 1–7. IEEE, 2006

    2006

  74. [81]

    Stratified round robin: A low complexity packet scheduler with bandwidth fairness and bounded delay

    Sriram Ramabhadran and Joseph Pasquale. Stratified round robin: A low complexity packet scheduler with bandwidth fairness and bounded delay. InProceedings of the 2003 conference on applications, technologies, architectures, and protocols for computer communi- cations, pages 23...

    2003

  75. [82]

    Diffspec: Differential testing with llms using natural language specifications and code artifacts.arXiv preprint arXiv:2410.04249, 2024

    Nikitha Rao, Elizabeth Gilbert, Harrison Green, Tahina Ramananandro, Nikhil Swamy, Claire Le Goues, and Sarah Fakhoury. Diffspec: Differential testing with llms using natural language specifications and code artifacts.arXiv preprint arXiv:2410.04249, 2024

    2024 arXiv

  76. [83]

    Swe-polybench: A multi-language benchmark for repository level evaluation of coding agents.arXiv preprint arXiv:2504.08703, 2025

    Muhammad Shihab Rashid, Christian Bock, Yuan Zhuang, Alexander Buchholz, Tim Esler, Si- mon Valentin, Luca Franceschi, Martin Wistuba, Prabhu Teja Sivaprasad, Woo Jung Kim, et al. Swe-polybench: A multi-language benchmark for repository level evaluation of coding agents.arXiv ...

    2025 arXiv

  77. [84]

    Hemem: Scalable tiered mem- ory management for big data applications and real nvm

    Amanda Raybuck, Tim Stamler, Wei Zhang, Mattan Erez, and Simon Peter. Hemem: Scalable tiered mem- ory management for big data applications and real nvm. InProceedings of the ACM SIGOPS 28th Sympo- sium on Operating Systems Principles, pages 392–407, 2021

    2021

  78. [85]

    Learning cache replacement with {CACHEUS}

    Liana V Rodriguez, Farzana Yusuf, Steven Lyons, Eysler Paz, Raju Rangaswami, Jason Liu, Ming Zhao, and Giri Narasimhan. Learning cache replacement with {CACHEUS}. In19th USENIX Conference on File and Storage Technologies (FAST 21), pages 341–

  79. [86]

    Mathematical discoveries from pro- gram search with large language models.Nature, 625(7995):468–475, 2024

    Bernardino Romera-Paredes, Mohammadamin Barekatain, Alexander Novikov, Matej Balog, M Pawan Kumar, Emilien Dupont, Francisco JR Ruiz, Jordan S Ellenberg, Pengming Wang, Omar Fawzi, et al. Mathematical discoveries from pro- gram search with large language models.Nature, 625(799...

    2024

  80. [87]

    Kerveros: Efficient and scalable cloud 19 admission control

    Sultan Mahmud Sajal, Luke Marshall, Beibin Li, Shan- dan Zhou, Abhisek Pan, Konstantina Mellou, Deepak Narayanan, Timothy Zhu, David Dion, Thomas Mosci- broda, et al. Kerveros: Efficient and scalable cloud 19 admission control. In17th USENIX Symposium on Operating Systems Desi...

    2023

  81. [88]

    How i learned to stop worrying and love learned os policies

    Divyanshu Saxena, Jiayi Chen, Sujay Yadalam, Yeonju Ro, Rohit Dwivedula, Eric H Campbell, Aditya Akella, Christopher J Rossbach, and Michael Swift. How i learned to stop worrying and love learned os policies. InProceedings of the 2025 Workshop on Hot Topics in Operating System...

    2025

  82. [89]

    {Self- Clocked}{Round-Robin} packet scheduling

    Erfan Sharafzadeh, Raymond Matson, Jean Tourril- hes, Puneet Sharma, and Soudeh Ghorbani. {Self- Clocked}{Round-Robin} packet scheduling. In22nd USENIX Symposium on Networked Systems Design and Implementation (NSDI 25), pages 1437–1465, Philadelphia, PA, 2025. USENIX Association

    2025

  83. [91]

    Reflexion: Language agents with verbal reinforcement learning

    Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. Reflexion: Language agents with verbal reinforcement learning. Advances in Neural Information Processing Systems, 36:8634–8652, 2023

    2023

  84. [92]

    Effi- cient fair queueing using deficit round robin

    Madhavapeddi Shreedhar and George Varghese. Effi- cient fair queueing using deficit round robin. InPro- ceedings of the conference on Applications, technolo- gies, architectures, and protocols for computer com- munication, pages 231–242, 1995

    1995

  85. [93]

    Code researcher: Deep research agent for large systems code and commit history

    Ramneet Singh, Sathvik Joel, Abhav Mehrotra, Nalin Wadhwa, Ramakrishna Bairi, Aditya Kanade, and Na- garajan Natarajan. Code researcher: Deep research agent for large systems code and commit history. Tech- nical Report MSR-TR-2025-34, Microsoft, June 2025

    2025

  86. [94]

    No silver bullet: ex- tending sdn to the data plane

    Anirudh Sivaraman, Keith Winstein, Suvinay Subra- manian, and Hari Balakrishnan. No silver bullet: ex- tending sdn to the data plane. InProceedings of the Twelfth ACM Workshop on Hot Topics in networks, pages 1–7, New York, NY , USA, 2013. Association for Computing Machinery

    2013

  87. [95]

    Learning relaxed belady for content distribu- tion network caching

    Zhenyu Song, Daniel S Berger, Kai Li, and Wyatt Lloyd. Learning relaxed belady for content distribu- tion network caching. In17th USENIX Symposium on Networked Systems Design and Implementation (NSDI 20), pages 529–544, Santa Clara, CA, 2020. USENIX Association

    2020

  88. [96]

    {HALP}: Heuristic aided learned preference eviction policy for {YouTube} content delivery net- work

    Zhenyu Song, Kevin Chen, Nikhil Sarda, Deniz Al- tınbüken, Eugene Brevdo, Jimmy Coleman, Xiao Ju, Pawel Jurczyk, Richard Schooler, and Ramki Gum- madi. {HALP}: Heuristic aided learned preference eviction policy for {YouTube} content delivery net- work. In20th USENIX Symposium ...

    2023

  89. [97]

    Xsbench-the development and verifi- cation of a performance abstraction for monte carlo reactor analysis.The Role of Reactor Physics toward a Sustainable Future (PHYSOR), 2014

    John R Tramm, Andrew R Siegel, Tanzima Islam, and Martin Schulz. Xsbench-the development and verifi- cation of a performance abstraction for monte carlo reactor analysis.The Role of Reactor Physics toward a Sustainable Future (PHYSOR), 2014

    2014

  90. [98]

    Speedy transactions in multicore in-memory databases

    Stephen Tu, Wenting Zheng, Eddie Kohler, Barbara Liskov, and Samuel Madden. Speedy transactions in multicore in-memory databases. InProceedings of the Twenty-Fourth ACM Symposium on Operating Systems Principles, pages 18–32, 2013

    2013

  91. [99]

    Evolution of the bfq storage-i/o scheduler

    Paolo Valente and Arianna Avanzini. Evolution of the bfq storage-i/o scheduler. In2015 Mobile Systems Technologies Workshop (MST), pages 15–20. IEEE, 2015

    2015

  92. [100]

    Automatic database management system tuning through large-scale machine learning

    Dana Van Aken, Andrew Pavlo, Geoffrey J Gordon, and Bohan Zhang. Automatic database management system tuning through large-scale machine learning. In Proceedings of the 2017 ACM international conference on management of data, pages 1009–1024, 2017

    2017

  93. [101]

    Andras Varga. Omnet++. InModeling and tools for network simulation, pages 35–59. Springer, 2010

    2010

  94. [102]

    Rodriguez, Wendy A

    Giuseppe Vietri, Liana V . Rodriguez, Wendy A. Mar- tinez, Steven Lyons, Jason Liu, Raju Rangaswami, Ming Zhao, and Giri Narasimhan. Driving cache re- placement with ML-based LeCaR. In10th USENIX Workshop on Hot Topics in Storage and File Systems (HotStorage 18), Boston, MA, J...

    2018

  95. [103]

    Efficient{MRC} construction with {SHARDS}

    Carl A Waldspurger, Nohhyun Park, Alexander Garth- waite, and Irfan Ahmad. Efficient{MRC} construction with {SHARDS}. In13th USENIX Conference on File and Storage Technologies (FAST 15), pages 95–110. USENIX Association, 2015

    2015

  96. [104]

    Pudica: Toward Near-Zero queuing delay in congestion control for cloud gaming

    Shibo Wang, Shusen Yang, Xiao Kong, Chenglei Wu, Longwei Jiang, Chenren Xu, Cong Zhao, Xuesong Yang, Jianjun Xiao, Xin Liu, Changxi Zheng, Jing Wang, and Honghao Liu. Pudica: Toward Near-Zero queuing delay in congestion control for cloud gaming. In21st USENIX Symposium on Netw...

    2024

  97. [105]

    Executable code actions elicit better llm agents

    Xingyao Wang, Yangyi Chen, Lifan Yuan, Yizhe Zhang, Yunzhu Li, Hao Peng, and Heng Ji. Executable code actions elicit better llm agents. InForty-first In- ternational Conference on Machine Learning, 2024

    2024

  98. [106]

    Chain-of-thought prompting elicits reasoning in large language models.Advances in neural information processing systems, 35:24824–24837, 2022

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. Chain-of-thought prompting elicits reasoning in large language models.Advances in neural information processing systems, 35:24824–24837, 2022

    2022

  99. [107]

    Multikernelbench: A multi-platform benchmark for kernel generation.arXiv eprints, pp

    Zhongzhen Wen, Yinghui Zhang, Zhong Li, Zhongxin Liu, Linna Xie, and Tian Zhang. Multikernelbench: A multi-platform benchmark for kernel generation.arXiv eprints, pp. arXiv–2507, 2025

    2025

  100. [108]

    Gandiva: Intro- spective cluster scheduling for deep learning

    Wencong Xiao, Romil Bhardwaj, Ramachandran Ram- jee, Muthian Sivathanu, Nipun Kwatra, Zhenhua Han, Pratyush Patel, Xuan Peng, Hanyu Zhao, Quanlu Zhang, Fan Yang, and Lidong Zhou. Gandiva: Intro- spective cluster scheduling for deep learning. In13th USENIX Symposium on Operatin...

    2018

  101. [109]

    Arms: Adaptive and robust memory tiering system.arXiv preprint arXiv:2508.04417, 2025

    Sujay Yadalam, Konstantinos Kanellis, Michael Swift, and Shivaram Venkataraman. Arms: Adaptive and robust memory tiering system.arXiv preprint arXiv:2508.04417, 2025

    2025 arXiv

  102. [110]

    Pantheon: the training ground for internet congestion- control research

    Francis Y Yan, Jestin Ma, Greg D Hill, Deepti Ragha- van, Riad S Wahby, Philip Levis, and Keith Winstein. Pantheon: the training ground for internet congestion- control research. In2018 USENIX Annual Technical Conference (USENIX ATC 18), pages 731–743, 2018

    2018

  103. [111]

    Swe-agent: Agent-computer interfaces enable automated software engineering.Advances in Neu- ral Information Processing Systems, 37:50528–50652, 2024

    John Yang, Carlos E Jimenez, Alexander Wettig, Kil- ian Lieret, Shunyu Yao, Karthik Narasimhan, and Ofir Press. Swe-agent: Agent-computer interfaces enable automated software engineering.Advances in Neu- ral Information Processing Systems, 37:50528–50652, 2024

    2024

  104. [112]

    Mithril: mining sporadic associations for cache prefetching

    Juncheng Yang, Reza Karimi, Trausti Sæmundsson, Avani Wildani, and Ymir Vigfusson. Mithril: mining sporadic associations for cache prefetching. InPro- ceedings of the 2017 Symposium on Cloud Computing, pages 66–79, 2017

    2017

  105. [113]

    Juncheng Yang, Ziming Mao, Yao Yue, and K. V . Rashmi. GL-Cache: Group-level learning for effi- cient and high-performance caching. In21st USENIX Conference on File and Storage Technologies (FAST 23), pages 115–134, Santa Clara, CA, February 2023. USENIX Association

    2023

  106. [114]

    Fifo can be better than lru: the power of lazy promotion and quick demotion

    Juncheng Yang, Ziyue Qiu, Yazhuo Zhang, Yao Yue, and KV Rashmi. Fifo can be better than lru: the power of lazy promotion and quick demotion. InProceed- ings of the 19th Workshop on Hot Topics in Operating Systems, pages 70–79, 2023

    2023

  107. [115]

    Fifo queues are all you need for cache eviction

    Juncheng Yang, Yazhuo Zhang, Ziyue Qiu, Yao Yue, and Rashmi Vinayak. Fifo queues are all you need for cache eviction. InProceedings of the 29th Sympo- sium on Operating Systems Principles, pages 130–149, New York, NY , USA, 2023. Association for Computing Machinery

    2023

  108. [116]

    {CacheSack}: Admission optimization for google datacenter flash caches

    Tzu-Wei Yang, Seth Pollen, Mustafa Uysal, Arif Merchant, and Homer Wolfmeister. {CacheSack}: Admission optimization for google datacenter flash caches. In2022 USENIX Annual Technical Confer- ence (USENIX ATC 22), pages 1021–1036, Carlsbad, CA, 2022. USENIX Association

    2022

  109. [117]

    Jonathan Chao

    Chen-Yu Yen, Soheil Abbasloo, and H. Jonathan Chao. Computers can learn from the heuristic designs and master internet congestion control. InProceedings of the ACM SIGCOMM 2023 Conference, ACM SIG- COMM ’23, page 255–274, New York, NY , USA, 2023. Association for Computing Machinery

    2023

  110. [118]

    An end-to- end automatic cloud database tuning system using deep reinforcement learning

    Ji Zhang, Yu Liu, Ke Zhou, Guoliang Li, Zhili Xiao, Bin Cheng, Jiashu Xing, Yangtao Wang, Tianheng Cheng, Li Liu, Minwei Ran, and Zekang Li. An end-to- end automatic cloud database tuning system using deep reinforcement learning. InProceedings of the 2019 In- ternational Confe...

    2019

  111. [119]

    Liteflow: towards high-performance adaptive neural networks for kernel datapath

    Junxue Zhang, Chaoliang Zeng, Hong Zhang, Shuihai Hu, and Kai Chen. Liteflow: towards high-performance adaptive neural networks for kernel datapath. InPro- ceedings of the ACM SIGCOMM 2022 Conference, SIGCOMM ’22, page 414–427, New York, NY , USA,

    2022

  112. [120]

    Yazhuo Zhang, Juncheng Yang, Yao Yue, Ymir Vig- fusson, and K.V . Rashmi. SIEVE is simpler than LRU: an efficient Turn-Key eviction algorithm for web caches. In21st USENIX Symposium on Networked Systems Design and Implementation (NSDI 24), pages 1229–1246, Santa Clara, CA, Apr...

    2024

  113. [121]

    Can llms replace time- tested system policies? perhaps

    Yibo Zhao and Cheng Tan. Can llms replace time- tested system policies? perhaps. InProceedings of the 16th ACM SIGOPS Asia-Pacific Workshop on Systems, pages 168–175, 2025. 21

    2025

  114. [122]

    Language agent tree search unifies reasoning acting and planning in language models.arXiv preprint arXiv:2310.04406, 2023

    Andy Zhou, Kai Yan, Michal Shlapentokh-Rothman, Haohan Wang, and Yu-Xiong Wang. Language agent tree search unifies reasoning acting and planning in language models.arXiv preprint arXiv:2310.04406, 2023

    2023 arXiv

  115. [123]

    3l-cache: Low overhead and precise learning-based eviction policy for caches

    Wenbin Zhou, Zhixiong Niu, Yongqiang Xiong, Juan Fang, and Qian Wang. 3l-cache: Low overhead and precise learning-based eviction policy for caches. In Proceedings of the 23rd USENIX Conference on File and Storage Technologies, pages 237–254, Santa Clara, CA, 2025. USENIX Association

    2025

  116. [124]

    The multi- queue replacement algorithm for second level buffer caches

    Yuanyuan Zhou, James Philbin, and Kai Li. The multi- queue replacement algorithm for second level buffer caches. InUSENIX Annual Technical Conference, Gen- eral Track, pages 91–104, 2001

    2001

  117. [125]

    Association for Computing Machinery

  118. [126]

    which machine gets the job?

    Tal Zussman, Ioannis Zarkadas, Jeremy Carin, Andrew Cheng, Hubertus Franke, Jonas Pfefferle, and Asaf Cidon. cache_ext: Customizing the page cache with ebpf. InProceedings of the ACM SIGOPS 31st Sympo- sium on Operating Systems Principles, SOSP ’25, page 462–478, New York, NY ...

    2025

  119. [131]

    Bigcodebench: Benchmarking code generation with diverse function calls and complex instructions.arXiv preprint arXiv:2406.15877, 2024

    Terry Yue Zhuo, Minh Chien Vu, Jenny Chim, Han Hu, Wenhao Yu, Ratnadira Widyasari, Imam Nur Bani Yusuf, Haolan Zhan, Junda He, Indraneil Paul, et al. Bigcodebench: Benchmarking code generation with diverse function calls and complex instructions.arXiv preprint arXiv:2406.15877, 2024

    2024 arXiv

  120. [133]

    template.h

    { 6int32 _t base _priority = obj _info.count * 20; 7int64 _t time _since_last_access = current _time - obj _info.last_access_vtime; 8base _priority -=static _cast<int32_t>(time_since_last_access / 300); 9base _priority -=static _cast<int32_t>(obj_info.size / 500); 10if(history...

  121. [354]

    USENIX Association, 2021

    2021

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.