Pith.

Artificial Intelligence · arXiv:2611.16487 v1 · submitted 12 Nov 2026 · 63 references

HeMeR: Heterogeneous Memory Reconciliation for Embodied Agents via Structured KV Reuse

Saehun Chun, Wonje Choi, Jinwoo Jang, Taeyoon Kwack, Honguk Woo

The paper says

“Embodied agents make decisions by combining information from heterogeneous memory modules, including declarative memory for explicit knowledge, procedural memory for learned routines, and working memory for current observations.”

more of the abstract

During interaction, the memory contents required for decision making change with the agent's state, goal, and task progress, while many contents remain shared across consecutive decisions. However, existing memory modules are typically accessed through separate interfaces, so each updated memory context is often processed as a new input to the foundation model. This prevents selective key-value (KV) cache reuse for unchanged memory contents and forces full KV cache recomputation even when much of the memory context remains valid. We introduce HeMeR, an inference framework that performs KV reconciliation between dynamically updated memory contexts and reusable KV states. HeMeR tracks declarative, procedural, and working memory contents across action decisions, preserves KV states for unchanged contents, and reconciles only the KV states affected by memory-content updates. This enables efficient and stable inference as memory contents evolve during interaction. We evaluate HeMeR on ALFWorld, Habitat, and real-world robotic deployments requiring coordinated use of heterogeneous memory contents. In the real-world setting, HeMeR achieves 83.20% task success and reduces per-step inference time from 2.94 to 1.87 seconds compared with MemRL, while preserving compatible KV states as task progress updates the memory context.

from the abstract

Pith says

Machine review · deepseek-v4-flash · 13 Nov 2026 · what was read was not recorded · not checked by a person

It can be wrong. Challenge this review or reply to it.

The passes disagreed. Reader: holds up. Skeptic: agrees with the reader. Referee: major revision.

Solid systems paper that frames heterogeneous memory updates as a KV reconciliation problem and validates it thoroughly, though the dependency annotation mechanism is underspecified. This line is wrong

1 major objection · 4 minor

How often objections from reviews like this one are right: not yet measured

Are you an author? What you can do.

Not checked: figures, code, what the 63 references say.

The paper says, at Section 4.3, paragraph 2; Algorithm 1, line 7

The review did not quote this passage.

Pith says: Objection 1 of 1 · major

The paper states that when the active adapter changes, HeMeR recomputes the procedural memory chunk's KV range and 'those of its direct and transitive dependents under the declared dependency annotations.' However, the paper never specifies how these dependency annotations are obtained—whether they are manually specified, automatically inferred from the foundation model's attention patterns, or derived from the memory schema's 'dependency annotation' field.

more of objection 1

The sensitivity analysis in Table 20 varies dependency density but does not validate that the actual annotations used in the experiments are correct or complete. If the annotations are incomplete, stale KV states could be reused, degrading task performance; if overapproximate, the method loses efficiency. This is a load-bearing gap because the correctness of the reconciliation plan (Equation 6 and Algorithm 1) depends on accurate dependency tracking. The authors should clarify the provenance of dependency annotations and provide evidence (e.g., human evaluation, automatic validation) that they are correct for the evaluated tasks.

4 minor objections

The paper says, at Table 3, caption

Pith says: minor objection 1 of 4

The caption says 'Analysis on cache-processing cost across reconciliation operations' but the table includes counts for events and tokens, not just costs. Consider rewording to 'Cache-processing cost decomposition by reconciliation operation'.

The paper says, at Section 5.2, paragraph 2

Pith says: minor objection 2 of 4

The sentence 'The efficiency and memory behavior metrics explain this trend' is vague. 'these metrics explain this trend' is imprecise; consider specifying which metrics (e.g., reuse ratio, active action ratio) explain the success gain.

show 2 more

The paper says, at Appendix B.2, pseudocode

Pith says: minor objection 3 of 4

The notation `use the same backbackground` is ambiguous: in Algorithm 1, line 13, the RoPE remapping formula should be written as `tilde{k}_j = R_theta(p'_j) R_theta(p_j)^{-1} k_j` for clarity, matching Equation (7). The current text uses inline notation but could b e more explicit.

The paper says, at Section 4.1, paragraph 3

Pith says: minor objection 4 of 4

The phrase 'Rather than treating the composed memory context Xt ... solely as a flat token sequence' is followed by 'HEMER organizes it as a sequence of memory chunks.' This is clear, but the subsequent sentence listing schema fields is a run-on; consider breaking it into a list or shorter sentences for readability.

What would settle it

Select one ALFWorld evaluation episode and manually enumerate for each memory-content update the set of chunks whose KV states must be recomputed because their attention can be affected (i.e., any chunk appearing after a changed chunk in the causal sequence). Compare this ground-truth set with the set actually recomputed by HeMeR's dependency annotations. If the recomputed set omits any chunk that should be recomputed, the dependency annotations are incomplete; if it includes extra chunks, they are overly broad. Report the precision and recall of the dependency propagation.

from the skeptic's reading

How the review went

4 readings by deepseek-v4-flash · 13 Nov 2026

  1. Reader holds up confidence high
  2. Skeptic disagrees 29 s · 34,870 tokens read, 3,265 written
  3. Editor wrote the verdict above 13 s · 35,612 tokens read, 1,340 written
  4. Referee major revision: 1 major objection, 4 minor 21 s · 35,136 tokens read, 2,303 written

Not run on this paper: the simulated authors' defence.

Pith did not check this review's objections against the paper.

The review in full

The review in brief, machine-written

This paper does a clean job of identifying a real problem: embodied agents with multiple memory modules recompute the full KV cache at every step because they treat updated memory contexts as completely new inputs. HeMeR’s framing of this as a KV reconciliation problem is the main conceptual contribution. The engineering integration – chunk-level indexing, RoPE remapping for position shifts, adapter-aware reconciliation for procedural switches – is sensible and put together in a way that the individual pieces alone don't cover.

The experiments are the paper’s strongest point. Three benchmarks, including real-world robot trials, with ablations and error bars. The cost decomposition (Table 3) shows that the majority of tokens are reused, and the ablation (Table 4) cleanly separates the contributions of each component. The hidden-state drift analysis (Figure 9) also gives reasonable confidence that the approximation doesn’t silently accumulate error. The authors acknowledge the shared-attention LoRA design limits cross-adapter drift but don’t eliminate it, and they quantify the drift (Table 18). That is fair.

The one thing that genuinely nags me is the dependency annotation mechanism described in Section 4.3. The paper says that when an adapter switches, HeMeR recomputes "the procedural memory chunk’s KV range and those of its direct and transitive dependents under the declared dependency annotations." But it never says how these dependencies are determined – manually written, inferred from attention patterns, or something else. The stress-test note on this is correct: if the annotations are incomplete, stale KV states might be reused; if they are over-approximate, the efficiency gain erodes. Table 20 varies dependency density for sensitivity but does not validate that the actual annotations used in the main experiments are correct. This is an underspecied load-bearing detail.

That said, the paper is otherwise well-executed and the central argument holds up. This is written for people building efficient serving stacks for memory-augmented robotics agents. It deserves a serious refereeing and should be accepted with a request to clarify the dependency annotation mechanism.

The strongest objection

HeMeR relies on declared dependency annotations to propagate recomputation from an updated memory chunk to its direct and transitive dependents (Section 4.3, paragraph 2). The paper does not describe how these dependencies are determined, whether they are manually specified, automatically inferred, or derived from the foundation model's attention pattern. If dependencies are incomplete, a dependent chunk might retain stale KV states after an update, degrading task performance. If they are overapproximate, unnecessary recomputation reduces efficiency. The sensitivity analysis in Table 20 varies dependency density but does not establish that the actual dependencies used in the experiments are correct or complete. Without this validation, the reconciliation plan's ability to localize updates while preserving correctness is undersupported.

The reader's weakest assumption was cross-adapter drift, which the paper acknowledges, controls via adapter-stable reconciliation, and bounds with ablation (SR from 71.43% to 64.85% when removed) and hidden-state drift analysis. That concern is adequately addressed. My concern targets a different, more foundational assumption: the dependency annotation mechanism that drives the reconciliation plan. The paper does not explain how dependencies are set, nor does it validate their correctness. This is load-bearing because incorrect dependencies would cause either stale KV reuse (hurting task success) or unnecessary recomputation (reducing efficiency gains). However, the paper's strong empirical results (Table 1, Table 2) and the sensitivity experiment (Table 20, varying dependency density from 0% to 100%) suggest that even under extreme densities the method performs well, partially mitigating the concern. Therefore I recommend no change to the verdict, but the concern deserves explicit future work or validation. I disagree with the reader's assessment of the weakest assumption because a more fundamental gap exists in the dependency specification.

The referee's report

If the results hold, HeMeR addresses a practically important problem: efficient KV cache reuse in memory-augmented embodied agents where memory contents change across decisions while many states remain reusable. The paper provides extensive empirical evaluation across three benchmarks, including real-world hardware, with ablations, sensitivity analyses, and representation-level drift characterization. The combination of schema-guided chunk indexing and adapter-aware reconciliation is novel relative to prior work on segment-level reuse (CacheBlend) and adapter switching (LRAgent). The reported efficiency gains (3-5x latency reduction) and task success improvements are substantial. The code and experimental details are provided, supporting reproducibility.

Pith's reading of the paper (a machine's summary, not the authors')

Embodied agents rely on declarative, procedural, and working memory modules that change across decisions. Current systems recompute the full key-value (KV) cache each time, wasting computation. HeMeR treats memory contents as typed chunks with associated KV ranges, maps updates to those ranges, and reuses or remaps unchanged states. On ALFWorld, Habitat, and real-world manipulation tasks, it achieves higher success rates and 2–5× lower per-step latency than strong baselines, without full cache rebuilds.

What carries the argument. Three components working together: (1) a memory schema that records each chunk's persistent ID, type, content hash, adapter identifier, and KV range; (2) chunk-level KV reconciliation that matches chunks across steps and reuses or recomputes based on schema compatibility, with RoPE remapping for position-only shifts; (3) adapter-stable KV reconciliation that includes the adapter identifier in the compatibility check and recomputes the procedural chunk's KV range when the adapter switches, while reusing other unaffected ranges.

Core claim. HeMeR establishes that heterogeneous memory updates—insertions, evictions, reorderings, and procedural-adapter switches—can be handled by reconciling only the affected KV ranges, preserving compatible states for unchanged content. The framework uses a memory schema to index chunks, RoPE-based position remapping for pure shifts, and adapter-stable reconciliation via shared-attention LoRA to limit cross-adapter drift. This enables efficient, long-horizon inference without full KV recomputation.

The premise it rests on. The shared-attention LoRA design limits but does not eliminate cross-adapter drift; the assumption that this drift is small enough to allow KV reuse across adapter switches is load-bearing.

If the paper is right (Pith's inferences, not the authors')

Enables embodied agents to operate over longer episodes with significantly less computational overhead, making them more practical for real-world deployment.

Provides a template for KV reuse in any interactive system where the input context evolves incrementally (e.g., conversational AI, code editing).

Reduces the gap between simulation and physical robot trials by lowering per-step latency to near-real-time levels.

Opens the possibility of concurrent KV reuse in multi-agent systems once concurrency control is added.

The schema-guided reconciliation approach could generalize beyond embodiment to any transformer-powered agent that maintains a mutable context (e.g., personal assistants that accumulate user preferences).

The paper's reliance on shared-attention LoRA to limit cross-adapter drift is a practical concession; if more diverse adapters are introduced, the assumption of small drift may need revisiting, perhaps through learned compatibility metrics.

The structured chunk representation hints at a principled cache-coherence protocol for LLM serving, analogous to invalidation-based coherence in distributed systems.

HeMeR's ablations suggest that even without adapter-stable reconciliation, reuse remains high but task success drops, indicating that representation quality matters more than raw cache hitrate.

Check this review

Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.

How often reviews like this one are right

Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.

How often objections from reviews like this one are right: not yet measured

Every value above, in the review's public record (JSON)

Reviews by people

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review

How Pith checks papers · The full audit of this paper

The paper

Abstract

Embodied agents make decisions by combining information from heterogeneous memory modules, including declarative memory for explicit knowledge, procedural memory for learned routines, and working memory for current observations. During interaction, the memory contents required for decision making change with the agent's state, goal, and task progress, while many contents remain shared across consecutive decisions. However, existing memory modules are typically accessed through separate interfaces, so each updated memory context is often processed as a new input to the foundation model. This prevents selective key-value (KV) cache reuse for unchanged memory contents and forces full KV cache recomputation even when much of the memory context remains valid. We introduce HeMeR, an inference framework that performs KV reconciliation between dynamically updated memory contexts and reusable KV states. HeMeR tracks declarative, procedural, and working memory contents across action decisions, preserves KV states for unchanged contents, and reconciles only the KV states affected by memory-content updates. This enables efficient and stable inference as memory contents evolve during interaction. We evaluate HeMeR on ALFWorld, Habitat, and real-world robotic deployments requiring coordinated use of heterogeneous memory contents. In the real-world setting, HeMeR achieves 83.20% task success and reduces per-step inference time from 2.94 to 1.87 seconds compared with MemRL, while preserving compatible KV states as task progress updates the memory context.

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.

abstract compiled · 2026-11-13

The authors' main claim, as the classifier picked it out 0.90

“In the real-world setting, HeMeR achieves 83.20% task success and reduces per-step inference time from 2.94 to 1.87 seconds compared with MemRL, while preserving compatible KV states as task progress updates the memory context.”

How the main claim is stated: asserted 1.00

What the abstract itself states, as probabilities: claims first 0.07 · states limitation 0.08 · human participants 0.07 · controlled comparison 0.88 · randomized assignment 0.01 · states quantitative result 0.99 · compares with prior results 0.97 · mentions code or data availability 0.03

The abstract, sentence by sentence

This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.

SentenceRoleThe authors' own workStates the main conclusionStrength
Embodied agents make decisions by combining information from heterogeneous memory modules, including declarative memory for explicit knowledge, procedural memory for learned routines, and working memory for current observations. background 1.00 0.07 0.02 asserted 1.00
During interaction, the memory contents required for decision making change with the agent's state, goal, and task progress, while many contents remain shared across consecutive decisions. background 1.00 0.11 0.02 asserted 1.00
However, existing memory modules are typically accessed through separate interfaces, so each updated memory context is often processed as a new input to the foundation model. gap 0.62 0.05 0.02 asserted 0.92
This prevents selective key-value (KV) cache reuse for unchanged memory contents and forces full KV cache recomputation even when much of the memory context remains valid. gap 0.58 0.08 0.03 asserted 0.99
We introduce HeMeR, an inference framework that performs KV reconciliation between dynamically updated memory contexts and reusable KV states. contribution 0.99 0.91 0.15 asserted 1.00
HeMeR tracks declarative, procedural, and working memory contents across action decisions, preserves KV states for unchanged contents, and reconciles only the KV states affected by memory-content updates. method 0.90 0.90 0.21 asserted 1.00
This enables efficient and stable inference as memory contents evolve during interaction. interpretation 0.76 0.39 0.20 asserted 0.98
We evaluate HeMeR on ALFWorld, Habitat, and real-world robotic deployments requiring coordinated use of heterogeneous memory contents. method 0.71 0.95 0.06 asserted 0.99
In the real-world setting, HeMeR achieves 83.20% task success and reduces per-step inference time from 2.94 to 1.87 seconds compared with MemRL, while preserving compatible KV states as task progress updates the memory context. result 1.00 0.97 0.48 asserted 1.00

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.99

What was examined, and what was not

Sources read

  • arXiv listing abstract v1

Not examined

  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers

Levels run

  • P abstract

Not run

  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

How the labels were made

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-13: 1 request, 6,963 input tokens.

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON

Figures (9)
Figure 1 of the paper
Figure 1. Motivation of HEMER. Environment changes can trigger memory-content updates, including procedural changes that switch the active adapter. A naive implementation recomputes the full KV cache, whereas HEMER updates affected KV ranges and reuses compatible KV states. current task state, including recent observations, intermediate decisions, and goal progress [2, 4, 17]. Although these modules are often studied through separate storage, retrieval, and update mechanisms, action selection requires their contents to be composed into a memory context. The resulting memory context is…
Figure 2 of the paper
Figure 2. HEMER overview. During long-horizon inference, new observations can modify the memory context and switch the active procedural adapter. HEMER indexes memory chunks, updates affected KV ranges, and preserves compatible cache states across decision steps. We distinguish three levels of memory representation: memory modules, memory contents, and memory contexts. A memory module stores and retrieves a specific type of information, and each item retrieved at a decision step is a memory content. A memory context is formed by composing multiple memory contents to support action selection.…
Figure 3 of the paper
Figure 3. Analysis on Cost efficiency and robustness under memory-content updates. (A)
Figure 4 of the paper
Figure 4. Example ALFWorld scenes. Environment. ALFWorld [32] is a text-based household environment in which an agent solves natural-language instructions through a discrete language action interface, as shown in
Figure 5 of the paper
Figure 5. Example EmbodiedBench Habitat scene and task observations.
Figure 6 of the paper
Figure 6. Example real-world environments. Task. Real-world tasks are executed using a compact set of predefined task-level skill APIs. As summarized in
Figure 7 of the paper
Figure 7. Real-world Dual-Section Kitchen-Café. The sequence includes two main subtasks: (1) making coffee with vanilla syrup and (2) making toast. – ReAct [2] interleaves reasoning traces and actions for step-by-step embodied decision making. – Reflexion [7] uses verbal self-reflection from previous failures to guide future action decisions. • Memory-augmented embodied baselines retrieve past experience or task-relevant memory to support long-horizon action selection. – MemRL [35] retrieves episodic memory from trajectories to improve embodied decision making. • KV cache reuse baselines…
Figure 9 of the paper
Figure 9. Analysis on Hidden-state drift from Full Rebuild shows that

Showing 8 of the paper's 9 figures. Figures are the authors', reproduced from the paper.

References (63)
  1. [1] Michael Ahn et al. “Do as i can, not as i say: Grounding language in robotic affordances”. In: arXiv preprint arXiv:2204.01691(2022) on PithPith's reviewarXiv
  2. [2] Shunyu Yao et al. “ReAct: Synergizing Reasoning and Acting in Language Models”. In:The Eleventh International Conference on Learning Representations (ICLR). 2023 on Pith
  3. [3] Chan Hee Song et al. “Llm-planner: Few-shot grounded planning for embodied agents with large language models”. In:Proceedings of the IEEE/CVF international conference on com- puter vision (ICCV). 2023 on Pith
  4. [4] Wenlong Huang et al. “Inner Monologue: Embodied Reasoning through Planning with Lan- guage Models”. In:6th Annual Conference on Robot Learning (CoRL). 2022 on Pith
  5. [5] Ishika Singh et al. “ProgPrompt: Program generation for situated robot task planning using large language models”. In:Autonomous Robots(2023) on Pith
  6. [6] Guanzhi Wang et al. “V oyager: An Open-Ended Embodied Agent with Large Language Models”. In:Transactions on Machine Learning Research (TMLR)(2024) on Pith
  7. [7] Noah Shinn et al. “Reflexion: language agents with verbal reinforcement learning”. In:Thirty- seventh Conference on Neural Information Processing Systems (NeurIPS). 2023 on Pith
  8. [8] Joon Sung Park et al. “Generative agents: Interactive simulacra of human behavior”. In: Proceedings of the 36th annual acm symposium on user interface software and technology. 2023 on Pith
  9. [9] Charles Packer et al. “MemGPT: Towards LLMs as Operating Systems”. In:ArXiv(2023) on Pith
  10. [10] Wanjun Zhong et al. “Memorybank: Enhancing large language models with long-term mem- ory”. In:Proceedings of the AAAI conference on artificial intelligence (AAAI). 2024 on Pith
  11. [11] Theodore Sumers et al. “Cognitive Architectures for Language Agents”. In:Transactions on Machine Learning Research (TMLR)(2024) on Pith
  12. [12] Wujiang Xu et al. “A-Mem: Agentic Memory for LLM Agents”. In:The Thirty-ninth Annual Conference on Neural Information Processing Systems (NeurIPS). 2025 on Pith
  13. [13] Wonje Choi et al. “NeSyC: A Neuro-symbolic Continual Learner for Complex Embodied Tasks in Open Domains”. In:Proceedings of the 13th International Conference on Learning Representations (ICLR 2025). 2025 on Pith
  14. [14] Neal J Cohen and Larry R Squire. “Preserved learning and retention of pattern-analyzing skill in amnesia: Dissociation of knowing how and knowing that”. In:Science(1980) on Pith
  15. [15] Wonje Choi, Jooyoung Kim, and Honguk Woo. “NeSyPr: Neurosymbolic Proceduralization for Efficient Embodied Reasoning”. In:Advances in Neural Information Processing Systems (NeurIPS 2025). 2025 on Pith
  16. [16] Edward J Hu et al. “LoRA: Low-Rank Adaptation of Large Language Models”. In:Interna- tional Conference on Learning Representations (ICLR). 2022 on Pith
  17. [17] Alan Baddeley. “Working memory”. In:Science(1992) on Pith
  18. [18] Reiner Pope et al. “Efficiently scaling transformer inference”. In:Proceedings of machine learning and systems (MLSys)(2023) on Pith
  19. [19] Woosuk Kwon et al. “Efficient Memory Management for Large Language Model Serving with PagedAttention”. In:Proceedings of the 29th Symposium on Operating Systems Principles (SOSP). 2023 on Pith
  20. [20] Endel Tulving et al. “Episodic and semantic memory”. In:Organization of memory(1972) on Pith
show 43 more
  1. [21] Jiayi Yao et al. “Cacheblend: Fast large language model serving for rag with cached knowledge fusion”. In:Proceedings of the twentieth European conference on computer systems (EuroSys). 2025. 11 on Pith
  2. [22] Jingbo Yang et al. “KVLink: Accelerating Large Language Models via Efficient KV Cache Reuse”. In:The Thirty-ninth Annual Conference on Neural Information Processing Systems (NeurIPS). 2025 on Pith
  3. [23] Yang Liu et al. “CacheSlide: Unlocking Cross Position-Aware KV Cache Reuse for Accelerat- ing LLM Serving”. In:24th USENIX Conference on File and Storage Technologies (FAST 26). USENIX Association, Feb. 2026 on Pith
  4. [24] Jianlin Su et al. “RoFormer: Enhanced transformer with Rotary Position Embedding”. In: Neurocomputing(2024) on Pith
  5. [25] Hyesung Jeon, Hyeongju Ha, and Jae-Joon Kim. “LRAgent: Efficient KV Cache Sharing for Multi-LoRA LLM Agents”. In:arXiv preprint arXiv:2602.01053(2026) on PithPith's reviewarXiv
  6. [26] Allison Li et al. “Efficient Multi-Adapter LLM Serving via Cross-Model KV-Cache Reuse with Activated LoRA”. In:arXiv preprint arXiv:2512.17910(2025) on PitharXiv
  7. [27] Zhiyu Li et al. “Memos: A memory os for ai system”. In:arXiv preprint arXiv:2507.03724 (2025) on PithPith's reviewarXiv
  8. [28] Ali Behrouz, Peilin Zhong, and Vahab Mirrokni. “Titans: Learning to Memorize at Test Time”. In:The Thirty-ninth Annual Conference on Neural Information Processing Systems (NeurIPS). 2025 on Pith
  9. [29] Junhao Hu et al. “EPIC: Efficient Position-Independent Context Caching for Serving Large Language Models”. In:Forty-second International Conference on Machine Learning (ICML). 2025 on Pith
  10. [30] Yuhan Liu et al. “CacheGen: KV Cache Compression and Streaming for Fast Large Language Model Serving”. In: ACM SIGCOMM ’24 (2024) on Pith
  11. [31] Yihua Cheng et al. “LMCache: An Efficient KV Cache Layer for Enterprise-Scale LLM Inference”. In:arXiv preprint arXiv:2510.09665(2025) on PitharXiv
  12. [32] Mohit Shridhar et al. “Alfworld: Aligning text and embodied environments for interactive learning”. In:arXiv preprint arXiv:2010.03768(2020) on PithPith's reviewarXiv
  13. [33] Rui Yang et al. “EmbodiedBench: Comprehensive Benchmarking Multi-modal Large Language Models for Vision-Driven Embodied Agents”. In:Forty-second International Conference on Machine Learning (ICML). 2025 on Pith
  14. [34] Andrew Szot et al. “Habitat 2.0: training home assistants to rearrange their habitat”. In: Proceedings of the 35th International Conference on Neural Information Processing Systems (NIPS). 2021 on Pith
  15. [35] Shengtao Zhang et al. “Memrl: Self-evolving agents via runtime reinforcement learning on episodic memory”. In:arXiv preprint arXiv:2601.03192(2026) on PithPith's reviewarXiv
  16. [36] Michał Zawalski et al. “Robotic Control via Embodied Chain-of-Thought Reasoning”. In:8th Annual Conference on Robot Learning (CoRL). 2024 on Pith
  17. [37] Andrew Zhao et al. “Expel: Llm agents are experiential learners”. In:Proceedings of the AAAI Conference on Artificial Intelligence (AAAI). 2024 on Pith
  18. [38] Zihao Wang et al. “ JARVIS-1: Open-World Multi-Task Agents With Memory-Augmented Multimodal Language Models ”. In:IEEE Transactions on Pattern Analysis & Machine Intelligence (TPAMI)(2025) on Pith
  19. [39] Runnan Fang et al. “Memp: Exploring Agent Procedural Memory”. In:arXiv preprint arXiv:2508.06433(2025) on PithPith's reviewarXiv
  20. [40] Minghao Chen et al. “AutoManual: Generating Instruction Manuals by LLM Agents via Interactive Environmental Learning”. In:The Thirty-eighth Annual Conference on Neural Information Processing Systems (NeurIPS). 2024 on Pith
  21. [41] Zhenhang Yuan, Shenghai Yuan, and Lihua Xie. “RPMS: Enhancing LLM-Based Embodied Planning through Rule-Augmented Memory Synergy”. In:arXiv preprint arXiv:2603.17831 (2026) on PitharXiv
  22. [42] Gyeong-In Yu et al. “Orca: A distributed serving system for Transformer-Based generative models”. In:16th USENIX symposium on operating systems design and implementation (OSDI 22). 2022 on Pith
  23. [43] In Gim et al. “Prompt cache: Modular attention reuse for low-latency inference”. In:Proceed- ings of Machine Learning and Systems (MLSys)(2024). 12 on Pith
  24. [44] Saehun Chun et al. “Functional Cache Grafting: Robust and Rapid Code-Policy Synthesis for Embodied Agents”. In:Proceedings of the 43rd International Conference on Machine Learning (ICML 2026). 2026 on Pith
  25. [45] Lianmin Zheng et al. “SGLang: efficient execution of structured language model programs”. In: Proceedings of the 38th International Conference on Neural Information Processing Systems (NIPS). 2024.ISBN: 9798331314385 on Pith
  26. [46] Zhekai Duan et al. “Fast ecot: Efficient embodied chain-of-thought via thoughts reuse”. In: arXiv preprint arXiv:2506.07639(2025) on PitharXiv
  27. [47] Zebin Yang et al. “KEEP: A KV-Cache-Centric Memory Management System for Efficient Embodied Planning”. In:arXiv preprint arXiv:2602.23592(2026) on PitharXiv
  28. [48] Zhenyu Zhang et al. “H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models”. In:Advances in Neural Information Processing Systems (NIPS). 2023 on Pith
  29. [49] Guangxuan Xiao et al. “Efficient Streaming Language Models with Attention Sinks”. In:The Twelfth International Conference on Learning Representations (ICLR). 2024 on Pith
  30. [50] Ying Sheng et al. “SLoRA: Scalable Serving of Thousands of LoRA Adapters”. In:Conference on Machine Learning and Systems (MLSys). 2024 on Pith
  31. [51] Lequn Chen et al. “Punica: Multi-tenant lora serving”. In:Proceedings of Machine Learning and Systems (MLSys). 2024 on Pith
  32. [52] Bingyang Wu et al. “dLoRA: Dynamically orchestrating requests and adapters for LoRA LLM serving”. In:18th USENIX Symposium on Operating Systems Design and Implementation (OSDI 24). 2024 on Pith
  33. [53] Hang Zhang et al. “Improving the serving performance of multi-lora large language models via efficient lora and kv cache management”. In:arXiv preprint arXiv:2505.03756(2025) on PithPith's reviewarXiv
  34. [54] AI@Meta.Llama 3.2 Model Card. 2024 on Pith
  35. [55] Shuai Bai et al. “Qwen3-vl technical report”. In:arXiv preprint arXiv:2511.21631(2025) on PithPith's reviewarXiv
  36. [56] Mohit Shridhar et al. “ALFRED: A Benchmark for Interpreting Grounded Instructions for Everyday Tasks”. In:Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). June 2020 on Pith
  37. [57] Homanga Bharadhwaj et al. “RoboAgent: Generalization and Efficiency in Robot Manipulation via Semantic Augmentations and Action Chunking”. In:2024 IEEE International Conference on Robotics and Automation (ICRA). 2024. 13 A Experiment setting A.1 ALFWorld (a) Bathroom scene exa... on Pith
  38. [58] take {obj} from {recep} on Pith
  39. [59] move {obj} to {recep} on Pith
  40. [60] clean {obj} with {recep} on Pith
  41. [61] heat {obj} with {recep} on Pith
  42. [62] cool {obj} with {recep} where {obj} and {recep} correspond to objects and receptacles. After each turn, the environment will provide feedback. If the environment outputs “Nothing happened”, the previous action is invalid and you should try another valid action. Your response s... on Pith
  43. [63] Institutional review board (IRB) approvals or equivalent for research with human subjects Question: Does the paper describe potential risks incurred by study participants, whether such risks were disclosed to the subjects, and whether Institutional Review Board (IRB) approvals... on Pith

Challenges and discussion

Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.

No comments yet.

Sign in to comment