PonderPounce: A Pretrained MLLM as an Episode Context Engine for Robot Control

• URL: http://arxiv.org/abs/2608.24115v1
• Date: Tue, 25 Aug 2026 06:24:36 GMT
• Title: PonderPounce: A Pretrained MLLM as an Episode Context Engine for Robot Control
• Abstract summary: PonderPounce reuses an MLLM's native causal context as robot memory.<n>Pounce receives the current observation, instruction, and proprioception directly.<n>On RoboMME with base-scale training data, PonderPounce reaches 60.83% with 9B and 50.04% with 0.8B under the same Pounce architecture and interface.
• Score: 25.920573627524945
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Multimodal large language models (MLLMs) can integrate long visual histories, reason under partial observability, and infer behavior from a few examples. Yet vision-language-action (VLA) models generally inherit pretrained representations without using this contextual capacity as episode memory. Memory-dependent policies address this gap through purpose-built history mechanisms. PonderPounce instead reuses an MLLM's native causal context as robot memory. Ponder, a System2 MLLM, accumulates episode observations, demonstrations, and prior cognition in its native causal context and can generate subgoal text and demonstration reasoning for internal use. Pounce, a System1 VLA, receives the current observation, instruction, and proprioception directly; through the Ponder--Pounce interface, it asynchronously receives only the newest continuous cognition token and its age. Both are jointly trained end to end without a purpose-built memory module or separate bridge pretraining. Optimized serving achieves p50 latencies of 78ms for cognition refresh and 25ms for action-model invocation, supporting 20Hz action playback. On RoboMME with base-scale training data, PonderPounce reaches 60.83% with 9B and 50.04% with 0.8B under the same Pounce architecture and interface, versus 44.51% for FrameSamp+Modul and 17.93% for the current-observation π_{0.5}. With 9x data, it reaches 75.54% versus 57.88% for FrameSamp+Modul. On RoboCasa-DC, the same interface learns from action supervision alone and reaches 12.5% versus 11.6% for the strongest published demonstration-conditioned baseline, falling to 8.6% when cognition is replaced by a learned null state.

Related papers

• GE-Act 2.0: Pretraining and Scaling a World-Action Model for Robotic Manipulation [63.853929983870955]
  We introduce Genie Envisioner Act 2.0 (GE-Act 2.0), a world-action model whose trainable generative and action components are all from scratch on manipulation data.<n>It combines a control-oriented autoencoder (CoAE), a single-step visual planner (SVP), and an inverse dynamics model (IDM)<n>The components are then jointly trained with knowledge-aligned selective optimization (KASO), which reduces mismatched supervision by selecting only predicted futures judged behaviorally compatible with the recorded action.
  arXiv  Detail & Related papers  (2026-09-04T17:16:48Z)
• CogEvol: Towards Efficient and Reliable Learning Environment Generation [49.18054685846454]
  We present CogEvol, a family of models trained specifically for Learning Environment Generation.<n>CogEvol completes a slide in a median of 17 seconds and an interactive page in 59, replacing minutes-long multi-turn agent scaffolding.<n>A production-grounded data pipeline turns real failures into 53,687 verified SFT samples.
  arXiv  Detail & Related papers  (2026-08-31T15:33:38Z)
• StreamPI: Streaming Multimodal Temporal Modeling for Vision-Language-Action Models [84.17128030384099]
  StreamPI is a streaming multimodal temporal modeling framework.<n>It equips single-frame VLA with temporal reasoning capability without introducing any additional parameters.<n> Experiments on real-robot tasks spanning memory-dependent and precise perception scenarios, as well as the simulation benchmark LIBERO, demonstrate that StreamPI outperforms pi0.5 across diverse tasks.
  arXiv  Detail & Related papers  (2026-08-26T17:33:19Z)
• AgentWeave: Routing Before Reasoning for Efficient Function Calling in Tool-Rich Language Models [0.0]
  We introduce AgentWeave, a deterministic pre-inference routing layer that constructs a bounded model-visible action space.<n>Relative to all-tools exposure, AgentWeave presents 70.18% fewer tools, uses 61.70% fewer input tokens, and exhibits 50.95% lower mean local-model latency.
  arXiv  Detail & Related papers  (2026-08-24T10:38:08Z)
• G0.5: One Autoregressive Stream for Robot Reasoning and Action [43.43377463558725]
  We introduce G0.5, a pretrained autoregressive VLA in which a single transformer decoder emits reasoning and action tokens under a single objective.<n>Because reasoning and action share a single set of weights, the pretrained VLM's capabilities carry over to physical behavior.<n>G0.5 surpasses state-of-the-art models across 7 independent regimes.
  arXiv  Detail & Related papers  (2026-08-12T07:26:47Z)
• Activity Frames: Deterministic Screen-Activity Compilation for Agent Memory and Replay [0.0]
  We compile passively captured screen activity into agent memory with a deterministic, zero-model pipeline.<n>It segments a local capture stream into typed activity frames, bounded episodes carrying application, site, timing, input volume, and evidence pointers back to the raw rows.<n>On one professional's single-user corpus of 128,756 frames over 51 active days, the compiler reduces a day of raw capture to a prompt-ready context block 86x smaller in 68 ms.
  arXiv  Detail & Related papers  (2026-08-06T09:17:43Z)
• ClawArena-Team: Benchmarking Subagent Orchestration and Dynamic Workflows in Language-Model Agents [61.184954205350984]
  ClawArena-Team is a benchmark of 41 multi-turn, multimodal, multi-agent scenarios spanning 258 evaluation rounds and 72 staged updates.<n>All scoring is execution-based with no LLM judge: an overall score -- the Subagent-Management Score (SMS) -- multiplies task correctness by a least-privilege and modality-routing factor.
  arXiv  Detail & Related papers  (2026-06-30T06:06:08Z)
• MemoryVAM: Integrating Memory into Video Action Model for Robot Manipulation [32.07039868974098]
  We present MemoryVAM, an episodic memory mechanism for video-world-model policies.<n>We employ a Recap-Cue (RC) module, in which a Perceiver-based Recap maps per-frame CLIP embeddings into compact memory tokens.<n>These tokens are injected into both the video backbone and the action decoder, aligning policy imagination with episode progress and conditioning actions on history.
  arXiv  Detail & Related papers  (2026-06-13T08:54:52Z)
• MEMTIER: Tiered Memory Architecture and Retrieval Bottleneck Analysis for Long-Running Autonomous AI Agents [6.8552591401084015]
  We present MEMTIER, a tripartite memory architecture for the OpenClaw agent.<n>On the full LongMemEval-S benchmark, MEMTIER achieves Acc.382, F1=.412 with Qwen2.5-7B on a consumer 6GB GPU.<n>With DeepSeek-V4-Flash fact pre-population, single-session recall reaches 0.686-0.714, exceeding the paper's RAG GPT-4o on those categories.
  arXiv  Detail & Related papers  (2026-05-05T12:14:10Z)
• Resource-Efficient Iterative LLM-Based NAS with Feedback Memory [49.44875022114861]
  Neural Architecture Search (NAS) automates network design, but conventional methods demand substantial computational resources.<n>We propose a closed-loop pipeline leveraging large language models (LLMs) to iteratively generate, evaluate, and refine convolutional neural network architectures.
  arXiv  Detail & Related papers  (2026-03-12T16:00:22Z)
• HAMLET: Switch your Vision-Language-Action Model into a History-Aware Policy [61.668591984635846]
  HAMLET is a framework to adapt Vision-Language-Action models to attend to the historical context during action prediction.<n>We show that HAMLET successfully transforms a state-of-the-art VLA into a history-aware policy.<n>On top of GR00T N1.5, HAMLET achieves an average success rate of 76.4% on history-dependent real-world tasks.
  arXiv  Detail & Related papers  (2025-10-01T09:15:52Z)
• MemoryVLA: Perceptual-Cognitive Memory in Vision-Language-Action Models for Robotic Manipulation [59.31354761628506]
  Temporal context is essential for robotic manipulation because such tasks are inherently non-Markovian, yet mainstream VLA models typically overlook it.<n>We propose MemoryVLA, a Cognition-Memory-Action framework for long-horizon robotic manipulation.<n>We evaluate it on 150+ simulation and real-world tasks across three robots.
  arXiv  Detail & Related papers  (2025-08-26T17:57:16Z)
• Seed1.5-Thinking: Advancing Superb Reasoning Models with Reinforcement Learning [231.11339402237903]
  We introduce Seed1.5-Thinking, capable of reasoning through thinking before responding.<n>Seed1.5-Thinking achieves 86.7 on AIME 2024, 55.0 on Codeforces and 77.3 on GPQA.<n>It demonstrates excellent reasoning abilities in STEM and coding.
  arXiv  Detail & Related papers  (2025-04-10T17:10:51Z)
• Perception Test: A Diagnostic Benchmark for Multimodal Video Models [78.64546291816117]
  We propose a novel multimodal video benchmark to evaluate the perception and reasoning skills of pre-trained multimodal models. The Perception Test focuses on skills (Memory, Abstraction, Physics, Semantics) and types of reasoning (descriptive, explanatory, predictive, counterfactual) across video, audio, and text modalities. The benchmark probes pre-trained models for their transfer capabilities, in a zero-shot / few-shot or limited finetuning regime.
  arXiv  Detail & Related papers  (2023-05-23T07:54:37Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.