Pith. sign in

REVIEW 5 major objections 7 minor 25 references

A behavior map earns its keep by telling a robot what to re-observe under a budget, not how to walk around a room.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-30 12:04 UTC pith:CJ76ERNW

load-bearing objection Solid reframe: behavior maps pay off as budgeted attention, not path costs; math and held-out evidence are clean, but the distinctive +2.5% motion-over-timestamp claim is thin and dynamics remain simulated. the 5 major comments →

arxiv 2607.23797 v1 pith:CJ76ERNW submitted 2026-07-26 cs.RO

Memory for Attention: Language-Conditioned Re-Perception with a Vision--Language--Motion Map

Dibyendu Ghosh This is my paper
classification cs.RO
keywords vision-language-motion mapsbudgeted re-perceptionattention schedulingopen-vocabulary mappingrestless banditsage of informationrobot memorylanguage-conditioned perception
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

A robot with a persistent map that records what moves faces two planning jobs. Shaping global paths around possible motion looks helpful on paper but nearly vanishes once the robot is actually driving and can replan reactively; an on-demand vision–language query does about as well. The job where memory matters is attention under a limited perception budget: which map entries to re-check so the representation stays fresh. History of actual change (or even simple recency) produces the best held-out schedule and matches an oracle, while a memoryless category prior is a poor scheduler. The gain equals the variance of root change-rates, concentrates on the objects that matter, and is largest when language names what to track—so open-vocabulary grounding and per-instance dynamics are both required.

Core claim

Under a limited perception budget, a persistent map’s memory of per-instance change yields the best re-perception schedule—matching an oracle and beating a memoryless on-demand VLM prior—with the memory-attributable gain equal to Var(√λ). In the language-conditioned setting, open-vocabulary relevance times observed dynamics beats even relevance-weighted recency and an on-demand VLM; neither language nor dynamics alone suffices. Spatial path-shaping benefit largely disappears in closed-loop execution.

What carries the argument

The √·-law attention schedule (re-check frequency proportional to √(w_i λ_i)) and its Cauchy–Schwarz consequence that the freshness gap versus a uniform/memoryless prior equals Var(√λ), the variance of root-volatility; a Whittle-index restless-bandit policy serves as the discrete reference that removes schedule artifacts.

Load-bearing premise

Object change is assumed to follow independent parametric processes whose per-instance heterogeneity can be dialed by a single knob; real long-term scene dynamics and natural heterogeneity are not measured.

What would settle it

Run the same held-out and language-conditioned schedules on a real robot with longitudinal, naturally occurring object moves and real detection success rates; if memory no longer beats recency and the on-demand prior once natural heterogeneity and viewpoint-coupled observation replace the simulated λ_i model, the central claim fails.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Map designers should treat behavior annotation primarily as an attention/resource-allocation signal, not as a global path-cost field.
  • When perception budget is scarce, storing per-instance change history (or at least last-seen times) is worth more than querying a category-level VLM on demand.
  • Language-conditioned tasks need both open-vocabulary grounding and a dynamics channel; a static open-vocabulary map plus recency is not enough.
  • The quantitative value of memory is predicted by measurable root-volatility variance, so scenes can be scored for whether a persistent map will pay off.
  • If observation reliability falls much below ~0.3 mean success, policies should fall back to the category prior rather than trust corrupted history.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same Var(√λ) logic suggests lifelong mapping systems should allocate mapping compute proportional to estimated root-volatility rather than uniform coverage sweeps.
  • Household and warehouse deployments with high instance-to-instance usage differences (same chair type used very differently) are the natural first places the claimed gain should appear at scale.
  • A practical system could gate the rate channel on measured observation reliability and importance skew, defaulting to cheapest recency otherwise.
  • The closed-loop path-shaping negative implies many ‘dynamic costmap’ papers may be over-claiming once reactive local control is admitted.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 7 minor

Summary. The manuscript contrasts two uses of a persistent Vision–Language–Motion Map (VLMM). For spatial navigation, behavior-aware path costs improve a planning-time objective but provide little closed-loop benefit. The main contribution reframes budgeted re-perception as attention: objects are assumed to change as point processes, a √-law allocates observation frequency, and a Cauchy–Schwarz argument predicts that the benefit of instance-level memory grows with root-volatility heterogeneity. Experiments in AI2-THOR with simulated change processes compare category priors, held-out change histories, oldest-first and Thompson schedulers, oracles, a Whittle-index policy, and a real CLIP prior. Memory generally performs best as heterogeneity increases, with robustness, reliability, fetch-task, and language-conditioned studies. The distinctive final claim is that language-conditioned VLMM scheduling beats relevance-weighted recency by +2.5% and an on-demand VLM by +8.9%, implying that both language grounding and per-instance dynamics are needed.

Significance. If the claims are appropriately qualified, the paper makes a useful conceptual contribution: persistent behavior maps may matter less for global path shaping than for deciding what to re-perceive. The closed-loop negative navigation result is valuable, and the attention study has several methodological strengths: transparent √-law and Cauchy–Schwarz arguments, a held-out rate-estimation protocol, explicit subtraction of the h=0 schedule floor, oracle and Whittle-index references, competitive age and Thompson baselines, process and observation-noise variants, real CLIP features, reliability reversal tests, and a falsifiable heterogeneity-scaling prediction. The work is therefore potentially influential for lifelong mapping and active perception, although its practical magnitude remains conditional on natural scene heterogeneity and observation reliability.

major comments (5)
  1. [§II, Eq. (1) and Proposition 1] The linear objective relies on λ_iτ_i≪1, but the default λ_max=0.12 and K=0.06N imply a mean revisit period near N/K≈16.7 and λτ as large as about 2. Prop. 1 also imposes only Σf_i=K, omitting the physical bounds 0≤f_i≤1 and the mapping to discrete top-K observations. The √-law therefore need not be a feasible optimum in the reported regime. Please quantify the surrogate error, impose or rule out binding box constraints, and compare with a schedule derived from the exact g(λτ).
  2. [§II, Corollary 1 and Eq. (2); §III-D, Fig. 5] Eq. (2) is derived for a homogeneous/global-mean memoryless baseline, whereas the experiments use category-varying priors ρ_i. At h=0, λ_i=λ_maxρ_i, so Var(√λ) is nonzero (reported as about 0.0009), yet the true memory-attributable gain is zero; Fig. 5's nonzero x-intercept makes the same point. Thus the headline claim that the gain “exactly equals Var(√λ)” is not the identity for the actual category prior. Please derive the Cauchy–Schwarz defect for λ̂=ρ_category, or use residual/within-category heterogeneity unexplained by that prior, and revise the abstract accordingly.
  3. [§III-F, Fig. 7] The paper's most distinctive claim—that the motion channel is more than a timestamp—rests on a +2.5% margin over relevance-weighted oldest-first, with t=7.4 but only 8 queries and 159 objects. The sampling unit and degrees of freedom are not stated, and objects are clustered by scene/category while sharing CLIP relevance and change histories; an object-level t-test may overstate effective sample size. Please report query- and scene-clustered confidence intervals or bootstrap tests, the per-query effects, and leave-one-query/scene-out results. Also specify how possibly negative CLIP cosine scores are transformed, since √(s_iλ̂_i) and s_i·age_i require nonnegative importance.
  4. [§III Setup, §III-C, and §IV(i–ii)] The central effect is demonstrated under generated λ_i=λ_max[(1−h)ρ_i+hu_i], while the natural value and dependence structure of h are unmeasured. The bursty, correlated, and diurnal variants are useful, but they remain mean-matched synthetic processes. Spatially or task-correlated shocks and stronger within-category homogeneity could materially shrink per-instance memory value. Please either estimate heterogeneity from real repeated scans/longitudinal data or add a hierarchical/correlated-shock sensitivity study and report gain against residual heterogeneity. At minimum, scope the conclusions explicitly as a simulation-established mechanism rather than a measured deployment effect.
  5. [§III-E, Table IV and Fig. 6; Abstract] The +21–26% result uses a zero-shot CLIP movability prior whose correlation with ground truth is only r=0.40. The manuscript appropriately notes that most of the base gain corrects this weak appearance prior, that a clean category prior beats memory below h=0.5, and that the base advantage shrinks as prior quality improves. The broader statements that an “on-demand VLM” is a poor scheduler are therefore too general. Please name the tested CLIP baseline in the abstract/conclusions, separate prior-correction gain from the heterogeneity slope, and ideally evaluate a stronger contextual VLM before making a general VLM claim.
minor comments (7)
  1. [§III-C, Eq. (3)] The symbol h denotes both per-instance heterogeneity and, in Eq. (3), the age-dependent holding cost h(τ). Rename the latter, e.g. c(τ), and avoid overloaded notation.
  2. [Fig. 1 and §II] Fig. 1 labels the budget B, while the text and Proposition 1 use K. Please unify the notation.
  3. [§III-F, Fig. 7] The exact query strings, number of independent runs/seeds, and unit underlying t=7.4 should be reported. Error bars are needed on the bars, and all five bar values should be legible.
  4. [Table III] The numerical definitions of “importance skew” 0/0.5/1/2, the log-normal importance parameters, and uncertainty intervals for each table entry are missing. These details are needed to reproduce the sensitivity conclusions.
  5. [§III-E] Please give the exact movable/fixed prompt pair, score calibration, and definition of the correlation r. This will help distinguish CLIP's movability estimate from a broader assessment of modern VLM capability.
  6. [§III-H, Table V] The reliability reversal is important, but only four mean reliability levels are shown. Please identify the detector, explain how measured detectability is scaled, report a finer curve with uncertainty, and state whether moving to a usable viewpoint consumes the same perception budget.
  7. [§III-I] The 35% planning-time objective, 4% closed-loop reduction, 14% travel penalty, and replan counts need formal metric definitions and preferably a table. This negative result is valuable and deserves the same reporting precision as the attention experiments.

Circularity Check

0 steps flagged

No significant circularity: scheduling identities follow from standard optimization; VLMM self-citation is substrate only, not load-bearing for the attention claims.

full rationale

The paper’s central mathematical chain (Prop. 1 √·-law, Prop. 2 Cauchy–Schwarz defect, Cor. 1 gap = Var(√λ)) is a standard continuous-surrogate derivation under an explicit Poisson/staleness model, not a fit renamed as a prediction. Heterogeneity h is an experimental control, not a parameter tuned to force the headline; held-out and Whittle protocols separate memory from schedule artifacts. Empirical checks (linear gap vs measured Var(√λ), R²=0.99; CLIP prior; fetch task; language ablations) are external to that identity. The only self-citation is VLMM [1] as the map representation that supplies open-vocabulary features and motion fields; the attention schedules, staleness metrics, and closed-loop negatives are independent AI2-THOR evaluations and do not reduce to unverified claims in [1]. Known √·-law / Whittle results are cited to external literature [19],[20], not smuggled from the authors’ prior ansatz. No self-definitional loop, no fitted-input-as-prediction, no uniqueness import. Score 1 only for the minor non-load-bearing substrate self-cite.

Axiom & Free-Parameter Ledger

5 free parameters · 6 axioms · 2 invented entities

Core math is standard optimization and bandits. Load-bearing domain assumptions are Poisson-like independent object change, budgeted top-K re-observation, and importance/relevance weights (including CLIP text alignment). Free knobs (h, λ_max, K, κ, w skew) control experiments rather than fit the central identity. Invented entities are mostly named constructs (VLMM fields, attention-as-reperception framing) built on the authors' prior map paper; independent evidence for VLMM as a map exists only via that citation, not new external validation here.

free parameters (5)
  • heterogeneity knob h = swept in [0,1]
    Controls mix of category prior vs per-instance uniform rates; central memory-attributable curves are reported vs h, whose real-world value is unmeasured.
  • λ_max = 0.12
    Global scale of change rates in simulation; sets absolute staleness level.
  • perception budget K = default 6% of N
    Fraction of objects re-observed per step; memory gain is budget-gated (near zero at 3%, larger at 20%).
  • Bayesian prior pseudocount κ = default order ~3; swept 1–30
    Strength of category prior in rate estimate λ̂_i=(m_i+κ ρ_i)/(t_i+κ); swept in sensitivity.
  • importance weights w_i = log-normal or CLIP similarity
    Log-normal draw or CLIP relevance s_i(q); skew changes concentration of benefit and memory vs oldest-first.
axioms (6)
  • domain assumption Each map element changes as an independent Poisson process (or bursty/correlated/diurnal variant) with fixed mean rate λ_i.
    Staleness formula g(λτ) and rate estimation rest on this; §II and process sweep §III-C.
  • standard math Value-weighted time-average staleness under continuous frequencies is minimized by f_i ∝ √(w_i λ_i) (Prop. 1 surrogate).
    Lagrange multiplier argument; paper notes discrete Whittle is the proper reference.
  • domain assumption On-demand VLM supplies only category-level movability ρ_i, not per-instance usage rates.
    Defines the memoryless baseline; real CLIP experiment uses appearance zero-shot movability.
  • domain assumption Language importance equals open-vocabulary feature alignment w_i = f_i^T CLIP_txt(q), independent of change rate.
    Language-conditioned experiments §III-F; measured corr(relevance, rate)≈0.
  • domain assumption Closed-loop low-level control and reactive replanning largely handle movers on the global path.
    Explains spatial-navigation negative §III-I; cited office-marathon style navigation robustness.
  • ad hoc to paper VLMM elements carry open-vocabulary features, movability prior ρ, motion score o, and confidence γ usable as map memory.
    Substrate from authors' prior arXiv:2607.16173; this paper evaluates downstream use rather than re-deriving the map.
invented entities (2)
  • Vision–Language–Motion Map (VLMM) as attention substrate no independent evidence
    purpose: Persistent open-vocabulary map with per-element change history driving re-perception schedules.
    Introduced in cited prior work [1]; here used as the memory-bearing map without new independent map-quality benchmarks.
  • Memory-attributable staleness reduction (gap net of h=0 schedule floor) independent evidence
    purpose: Isolate true value of per-instance history from √·-law discrete suboptimality.
    Analysis construct defined in §III setup; Whittle section later removes the floor more cleanly.

pith-pipeline@v1.2.0-grok45-kimik3 · 17610 in / 4061 out tokens · 76129 ms · 2026-07-30T12:04:54.656962+00:00 · methodology

0 comments
read the original abstract

A robot carrying a persistent, behavior-annotated map faces two planning questions, and its memory answers only one well. The \emph{spatial-navigation} question -- how to walk around a room -- we address first and report a negative: building on Vision--Language--Motion Maps (VLMM), a behavior-aware planner cost cuts a planning-time objective by $\sim$35\% over 28 AI2-THOR scenes, but under closed-loop execution the real benefit nearly vanishes ($\sim$4\%) and an on-demand vision--language model (VLM) does as well. The \emph{resource-allocation} question differs: under a limited perception budget, what should the robot re-observe now to keep its map fresh? Framing re-perception as this attention decision, we show a persistent map's memory (change-history, or even just recency of last sighting) yields the best schedule (held-out), matching an oracle, while the memoryless VLM prior is poor. Because the schedule reallocates budget toward what matters, memory's benefit concentrates on the important objects ($\sim$1.6$\times$ the mean), and a downstream fetch task confirms fewer wasted trips; the gain grows with per-instance heterogeneity exactly as a Cauchy--Schwarz bound predicts -- it equals $\mathrm{Var}(\sqrt\lambda)$, the variance of root-volatility. With a real CLIP prior on rendered objects the advantage is $+21$--$26\%$. The map's distinctive value appears when the task is \emph{language-conditioned}: told what to track, VLMM grounds the relevant objects (open-vocabulary) and tracks their change (memory), beating even a strong relevance-weighted recency baseline ($+2.5\%$) -- so its motion channel adds value beyond a last-seen timestamp -- and an on-demand VLM ($+8.9\%$); neither language nor dynamics alone suffices. The map earns its keep not by telling the robot how to walk around a room, but by telling it what to pay attention to.

Figures

Figures reproduced from arXiv: 2607.23797 by Dibyendu Ghosh.

Figure 1
Figure 1. Figure 1: Budgeted re-perception as attention. The map holds each element’s change history; a rate estimate λˆ i is formed from the on-demand category prior (VLM) or the map’s observed history (memory). The √ ·-law schedule (Prop. 1) sets each element’s re-check frequency, a budget of K selects which to re-observe, and each observation updates the history (loop); we score value-weighted staleness on a held-out reali… view at source ↗
Figure 2
Figure 2. Figure 2: Held-out attention policy, 951 objects. As per-instance rates deviate from the category prior (increasing h), observed history (memory, teal) pulls value-weighted staleness below the on-demand-VLM prior (red). The h=0 gap is a schedule-level floor, not memory; the memory-attributable effect is the heterogeneity-driven separation. the comparison isolates the rate estimate, not a deployable system. We add th… view at source ↗
Figure 3
Figure 3. Figure 3: Competitive schedulers, run online (skewed importance). The on [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Memory-attributable staleness reduction (held-out, net of the [PITH_FULL_IMAGE:figures/full_fig_p004_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Prior-quality sweep. As a category prior’s accuracy [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗
Figure 5
Figure 5. Figure 5: Empirical test of Cor. 1. The realised held-out memory reduction is [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 7
Figure 7. Figure 7: Language-conditioned re-perception (online arena, relevance [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

25 extracted references · 13 linked inside Pith

  1. [1]

    Vision-Language-Motion Maps: An open- vocabulary, uncertainty-aware, queryable motion attribute for 3D scene maps,

    D. Ghosh and A. Shakya, “Vision-Language-Motion Maps: An open- vocabulary, uncertainty-aware, queryable motion attribute for 3D scene maps,”arXiv preprint arXiv:2607.16173, 2026

    Pith/arXiv arXiv 2026

  2. [2]

    Visual language maps for robot navigation,

    C. Huang, O. Mees, A. Zeng, and W. Burgard, “Visual language maps for robot navigation,” inIEEE International Conference on Robotics and Automation (ICRA), 2023, arXiv:2210.05714

    Pith/arXiv arXiv 2023

  3. [3]

    Conceptfusion: Open-set multimodal 3d mapping,

    K. M. Jatavallabhulaet al., “Conceptfusion: Open-set multimodal 3d mapping,” inRobotics: Science and Systems (RSS), 2023, arXiv:2302.07241

    Pith/arXiv arXiv 2023

  4. [4]

    Conceptgraphs: Open-vocabulary 3d scene graphs for perception and planning,

    Q. Guet al., “Conceptgraphs: Open-vocabulary 3d scene graphs for perception and planning,” inIEEE International Conference on Robotics and Automation (ICRA), 2024, arXiv:2309.16650

    Pith/arXiv arXiv 2024

  5. [5]

    Hierarchical open-vocabulary 3d scene graphs for language-grounded robot navigation,

    A. Werby, C. Huang, M. B ¨uchner, A. Valada, and W. Burgard, “Hierarchical open-vocabulary 3d scene graphs for language-grounded robot navigation,” inRobotics: Science and Systems (RSS), 2024, arXiv:2403.17846

    Pith/arXiv arXiv 2024

  6. [6]

    Khronos: A unified approach for spatio-temporal metric-semantic slam in dynamic environments,

    L. Schmidet al., “Khronos: A unified approach for spatio-temporal metric-semantic slam in dynamic environments,” inRobotics: Science and Systems (RSS), 2024, arXiv:2402.13817

    Pith/arXiv arXiv 2024

  7. [7]

    Deep semantic classi- fication for 3d lidar data,

    A. Dewan, G. L. Oliveira, and W. Burgard, “Deep semantic classi- fication for 3d lidar data,” inIEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017, arXiv:1706.08355

    Pith/arXiv arXiv 2017

  8. [8]

    The office marathon: Robust navigation in an indoor office environ- ment,

    E. Marder-Eppstein, E. Berger, T. Foote, B. Gerkey, and K. Konolige, “The office marathon: Robust navigation in an indoor office environ- ment,” inIEEE Int. Conf. on Robotics and Automation (ICRA), 2010, pp. 300–307

    2010

  9. [9]

    Dynamem: Online dynamic spatio-semantic memory for open world mobile manipulation,

    P. Liu, Z. Guo, M. Warke, S. Chintala, C. Paxton, N. M. M. Shafiullah, and L. Pinto, “Dynamem: Online dynamic spatio-semantic memory for open world mobile manipulation,”arXiv preprint arXiv:2411.04999, 2024

    Pith/arXiv arXiv 2024

  10. [10]

    Dynamic open-vocabulary 3d scene graphs for long-term language- guided mobile manipulation,

    Z. Yan, S. Li, Z. Wang, L. Wu, H. Wang, J. Zhu, L. Chen, and J. Liu, “Dynamic open-vocabulary 3d scene graphs for long-term language- guided mobile manipulation,”arXiv preprint arXiv:2410.11989, 2025

    Pith/arXiv arXiv 2025

  11. [11]

    Dualmap: Online open- vocabulary semantic mapping for natural language navigation in dynamic changing scenes,

    J. Jiang, Y . Zhu, Z. Wu, and J. Song, “Dualmap: Online open- vocabulary semantic mapping for natural language navigation in dynamic changing scenes,”arXiv preprint arXiv:2506.01950, 2025

    arXiv 2025

  12. [12]

    Model- ing dynamic environments with scene graph memory,

    A. Kurenkov, M. Lingelbach, T. Agarwal, E. Jin, C. Li, R. Zhang, L. Fei-Fei, J. Wu, S. Savarese, and R. Mart ´ın-Mart´ın, “Model- ing dynamic environments with scene graph memory,” inInterna- tional Conference on Machine Learning (ICML), vol. 202, 2023, arXiv:2305.17537

    Pith/arXiv arXiv 2023

  13. [13]

    Active perception,

    R. Bajcsy, “Active perception,”Proceedings of the IEEE, vol. 76, no. 8, pp. 966–1005, 1988

    1988

  14. [14]

    Revisiting active per- ception,

    R. Bajcsy, Y . Aloimonos, and J. K. Tsotsos, “Revisiting active per- ception,”Autonomous Robots, vol. 42, no. 2, pp. 177–196, 2018

    2018

  15. [15]

    The determination of next best views,

    C. I. Connolly, “The determination of next best views,” inIEEE International Conference on Robotics and Automation (ICRA), vol. 2, 1985, pp. 432–435

    1985

  16. [16]

    Receding horizon “next-best-view

    A. Bircher, M. Kamel, K. Alexis, H. Oleynikova, and R. Siegwart, “Receding horizon “next-best-view” planner for 3D exploration,” in IEEE International Conference on Robotics and Automation (ICRA), 2016, pp. 1462–1468

    2016

  17. [17]

    Persistent robotic tasks: Mon- itoring and sweeping in changing environments,

    S. L. Smith, M. Schwager, and D. Rus, “Persistent robotic tasks: Mon- itoring and sweeping in changing environments,”IEEE Transactions on Robotics, vol. 28, no. 2, pp. 410–426, 2012

    2012

  18. [18]

    Persistent monitoring in discrete environments: Minimizing the maximum weighted latency be- tween observations,

    S. Alamdari, E. Fata, and S. L. Smith, “Persistent monitoring in discrete environments: Minimizing the maximum weighted latency be- tween observations,”The International Journal of Robotics Research, vol. 33, no. 1, pp. 138–154, 2014

    2014

  19. [19]

    Effective page refresh policies for web crawlers,

    J. Cho and H. Garcia-Molina, “Effective page refresh policies for web crawlers,”ACM Transactions on Database Systems, vol. 28, no. 4, pp. 390–426, 2003

    2003

  20. [20]

    Restless bandits: Activity allocation in a changing world,

    P. Whittle, “Restless bandits: Activity allocation in a changing world,” Journal of Applied Probability, vol. 25A, pp. 287–298, 1988

    1988

  21. [21]

    Scheduling policies for minimizing age of information in broadcast wireless networks,

    I. Kadota, A. Sinha, E. Uysal-Biyikoglu, R. Singh, and E. Modiano, “Scheduling policies for minimizing age of information in broadcast wireless networks,”IEEE/ACM Transactions on Networking, vol. 26, no. 6, pp. 2637–2650, 2018

    2018

  22. [22]

    Age of information: An introduction and survey,

    R. D. Yates, Y . Sun, D. R. Brown, S. K. Kaul, E. Modiano, and S. Ulukus, “Age of information: An introduction and survey,”IEEE Journal on Selected Areas in Communications, vol. 39, no. 5, pp. 1183–1210, 2021

    2021

  23. [23]

    Change rate estimation and optimal freshness in web page crawling,

    K. Avrachenkov, K. Patil, and G. Thoppe, “Change rate estimation and optimal freshness in web page crawling,” in13th EAI International Conference on Performance Evaluation Methodologies and Tools (VALUETOOLS), 2020, arXiv:2004.02167

    Pith/arXiv arXiv 2020

  24. [24]

    AI2-THOR: An interactive 3d environment for visual ai,

    E. Kolveet al., “AI2-THOR: An interactive 3d environment for visual ai,”arXiv preprint arXiv:1712.05474, 2017

    Pith/arXiv arXiv 2017

  25. [25]

    Learning transferable visual models from natural language super- vision,

    A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell, P. Mishkin, J. Clark, G. Krueger, and I. Sutskever, “Learning transferable visual models from natural language super- vision,” inInternational Conference on Machine Learning (ICML), 2021, arXiv:2103.00020

    Pith/arXiv arXiv 2021