Searcharxiv⌕ Search

arXiv subjects

Pulkit Agrawal

Publications and source records attributed to Pulkit Agrawal.

At least 19 recordsLinked to original sources
arXiv2026

When Robots Outweigh Objects: How Inertia Mismatch Shapes Robot Policy Learning

Conventional robot arms can be extremely heavy compared with the everyday objects they manipulate, creating a significant inertia mismatch at contact. Even a small accidental bump from a heavy arm can knock a light object out of place. Modern learning methods have nevertheless demonstrated a wide range of successful manipulation behaviors on these platforms, suggesting that even extreme inertia mismatch can be overcome with sufficient data. However, it remains unclear whether high inertia imposes a hidden cost on learning, potentially worsening scaling with data or lowering the performance ceiling of modern robot learning methods. We investigate this question through controlled simulation experiments that vary robot and object masses. We find that reinforcement learning succeeds across a broad range of mass combinations, while behavior cloning learns more reliable policies when a lighter robot manipulates a heavier object, and more data narrows the resulting gap without closing it. Behavior cloning likewise degrades on a simulated robot arm as its motors reflect more inertia through their gearing, even though its weight stays the same. Finally, because new hardware is rarely an option, we explore ways to lower contact inertia on the arm one already has, either by choosing a lower-inertia workspace before collecting data or by steering the demonstrator toward lower-inertia solutions of the task. In both simulated cases, the lower-inertia demonstrations are imitated more reliably. Project website: https://younghyopark.me/inertia-mismatch/

cs.RO↗
arXiv2026

Conditional Flow Models Convert Epistemic Uncertainty into Apparent Aleatoric Variance

Conditional Flow Models (CFMs) are increasingly being used in place of deterministic predictors across machine learning, including for tasks that were classically formulated as regression. Given their success in such tasks, we ask the question: what useful information do the conditional distributions produced by CFMs carry beyond what a point predictor would provide? We identify a systematic phenomenon: in many natural regimes, CFM samples routinely have larger variance than the true conditional distribution, even in nearly deterministic regression problems. This effect is most pronounced where the model has not resolved the dependence on the conditioning variable, and thus the additional variance can be viewed as converting epistemic uncertainty in fitting the ground truth conditional distribution into apparent aleatoric variance in the generated samples. Through controlled synthetic and real-data experiments, and comparisons with explicit smoothing estimators, we show that the induced variance tracks model underfitting of complex or hard-to-fit targets, and decreases with additional training or data. We provide a theoretical account of the phenomenon in a simplified setting and demonstrate a consequent practical use-case of conditional flow-matching: designing high-coverage predictive sets that cover the ground-truth regression target.

cs.LG↗
arXiv2026

Blind Precise Assembly: Proprioceptive Sim-to-Real Peg-in-Hole for Diverse Parts

Every robot already carries high-resolution contact sensors: its own joints. Legged robots and humanoids exploit them to walk, run, and track whole-body motions without vision, but there the contacts are large and forceful. Is joint sensing also enough for precise, contact-rich assembly, where the robot must find a millimeter-scale opening, mesh gear teeth, or catch a thread through light contact? We study this question by training blind assembly policies that observe only a short history of joint positions, velocities, and their own actions, without visual, force, torque, or tactile sensing. The poses of both the fixture and the part are unknown to the policy, and the grasp is randomized substantially. The policy must therefore assemble the parts through contact, sensed exclusively through its own proprioception. Trained with reinforcement learning entirely in simulation over randomized parts, grasps, and contacts, such policies learn peg insertion, gear meshing, and nut threading. The peg insertion policy transfers zero-shot to the real world and inserts pegs of different shapes and sizes in 73% of trials. Controlled comparisons with matched visual policies in simulation show that blind policies generalize far better to unseen part shapes and sizes, though they can be slower. Training such policies needs thousands of different assemblies, each simulated with its clearance intact, so we also release an open-source generator that builds them procedurally from exact convex pieces for MuJoCo; it is more accurate and faster than convex decomposition or MuJoCo's signed-distance collision. Videos and interactive simulations are at https://younghyopark.me/blind-proprioceptive-assembly/.

cs.RO↗
arXiv2026

Compliant GMT: Making General Motion Tracking Compliant via Residual Reference Adaptation

We propose Compliant GMT (C-GMT), a modular framework for full-body compliant responses in general motion tracking. As humanoid robots move into real world deployment, they not only need to perform any motion but also respond safely and robustly to interaction forces. Existing general motion tracking policies can imitate diverse motions but are trained to be stiff, leading to unexpected and unsafe behaviors under disturbance. Prior work in compliant humanoid control trades off compliance and generality. There exist examples of full-body compliant responses, but these are heavily curated to single-reference policies. On the other hand, general motion trackers (GMT) are trained to rigidly imitate massive motion capture datasets. Our key insight is to leverage the generality of pretrained GMTs for motion tracking while introducing compliance from a residual adapter that conditions on a history of proprioception. This enables our approach, C-GMT, to achieve both general motion tracking and whole-body compliant responses. Compared to state-of-the-art general motion trackers and single-motion compliant policies, C-GMT demonstrates competitive performance in both motion tracking accuracy and compliance adherence. We further demonstrate improved safety, robustness, and error tolerance in simulation and real-world G1 experiments.

cs.RO↗
arXiv2026

Contact-UAN: Unsupervised Actuator Networks as a Residual World Model for Humanoid Control

Humanoid policies trained in simulation can transfer qualitatively to hardware while failing to transfer the force interactions learned during training, compromising torque efficiency, load regulation, and robustness to impacts. This failure can arise from actuator sim-to-real gaps because forces depend on actuator torques. We introduce Contact-UAN, a residual world model that preserves a simulator's physics priors while correcting actuator dynamics to align simulated rollouts with recorded hardware rollouts. For humanoid motion with contact, this requires contact-consistent replay: if hardware playback in simulation begins with misaligned contact wrenches, the residual model learns to correct reset artifacts rather than actuator discrepancies. Using only onboard proprioception, Contact-UAN infers the simulator-side stance-foot pose needed to initialize replay with consistent contact forces. We show that a UAN trained on 6.5 minutes of real-world walking data generalizes to tasks that benefit from accurate force and impulse regulation and produces policies that transfer seamlessly to the real world, yielding 20% lower running torque, 43% lower force-tracking error, extreme compliance, accurate impedance control, and dynamic, robust jumps.

cs.RO↗
arXiv2026

Large Pretraining Datasets Don't Guarantee Robustness after Fine-Tuning in Image Classification

Large-scale pretrained models are widely leveraged as foundations for learning new specialized tasks via fine-tuning, with the goal of maintaining the general performance of the model while allowing it to gain new skills. A valuable goal for all such models is robustness: the ability to perform well on out-of-distribution (OOD) tasks. We assess whether fine-tuning preserves the overall robustness of the pretrained model in image classification, and observed that models pretrained on large datasets exhibited strong catastrophic forgetting and loss of OOD generalization. To systematically assess robustness preservation in fine-tuned models, we propose the Robustness Inheritance Benchmark (ImageNet-RIB). The benchmark, which can be applied to any pretrained model, consists of a set of related but distinct OOD (downstream) tasks and involves fine-tuning on one of the OOD tasks in the set then testing on the rest. We find that though continual learning methods help, fine-tuning reduces robustness across pretrained models. Surprisingly, models pretrained on the largest and most diverse datasets (e.g., LAION-2B) exhibit both larger robustness losses and lower absolute robustness after fine-tuning on small datasets, relative to models pretrained on smaller datasets. We observe this collapse in contrastively pretrained (CLIP) models and their fine-tuned variants, where it grows with pretraining scale; the supervised models we test do not exhibit it. These findings suggest that starting with the strongest foundation model is not necessarily the best approach for performance on specialist tasks. https://jd730.github.io/projects/ImageNet-RIB

cs.CV↗
arXiv2026

Learning from Hindsight for VLA Reinforcement Learning

Reinforcement learning is increasingly used to fine-tune vision-language-action (VLA) models, but robot interaction is expensive and learning becomes highly sample inefficient when successful rollouts are rare. When reward is assigned only for completing the commanded task, a failed rollout is treated as having no value even if it successfully executes behaviors relevant to that task. A robot that fails to place the correct object in a bowl may still move that object toward the bowl or place a different object inside it, demonstrating objects and actions that can be reused to solve the target task. These behaviors define auxiliary tasks that the policy can already solve, providing useful learning signals even before it can solve the harder target task. We introduce $\textit{Learning from Hindsight (LfH)}$, which turns such failures into additional learning signals. Using a pretrained vision-language model, LfH relabels failed rollouts with the behaviors they actually accomplish and trains the policy jointly on the commanded task and these auxiliary tasks. On out-of-distribution LIBERO-PRO manipulation tasks, LfH matches the final performance of GRPO with approximately $5\times$ fewer rollouts and improves sample efficiency across multiple VLA backbones. On a physical Franka robot, LfH raises success from $0\%$ to $56\%$ within 160 training rollouts, while GRPO reaches $22\%$.

cs.LG↗
arXiv2026

Workspace Models: Lightweight Robotic Memory via Saliency-Driven Supervision

Complex robotic manipulation tasks frequently require a long-term memory of past events and actions. As conditioning on full histories renders policies prone to spurious correlations and degrades performance, many approaches to policy memory involve compressing historical information through expensive VLM queries in-the-loop to process only task-salient information. In this paper, we propose an alternative approach in which computationally intensive VLM queries are made during train-time to learn a lightweight latent memory that can be efficiently queried at deployment time. Our representation, which we call the workspace token, is trained by (1) using a VLM to identify current and historical information necessary for completing a task, then (2) distilling these into the workspace token using a set-reconstruction decoder loss. In both simulation and hardware, we show that the workspace token can be used as a drop-in replacement for observations during deployment, enabling policies to solve memory-intensive tasks without the need for VLM reasoning in-the-loop, in effect serving as a latent harness for distilling a stronger reasoning models ability to solve long-horizon tasks to a reactive robotic policy. We further demonstrate that the workspace tokens are not only more lightweight, but also lead to better policy performance compared to conditioning policies on explicit modalities like curated past image frames, motivating a latent approach to history curation and reasoning model harnesses more broadly.

cs.RO↗
arXiv2026

DexWrist: A Robotic Wrist for Constrained and Dynamic Manipulation

Development of dexterous manipulation hardware has primarily focused on hands and grippers. However, these end-effectors are often paired with bulky and highly stiff wrists that limit performance in human environments. More recent designs have adopted backdrivable actuation, but are still difficult to model and control due to coupled kinematics or high mechanical inertia from heavy links. We present DexWrist, a compact robotic wrist combining quasi-direct-drive actuation with a decoupled parallel kinematic mechanism to advance manipulation in highly constrained environments and enable dynamic, contact-rich tasks. It delivers 3.75$\pm$0.05 Nm rated torque, 0.33$\pm$0.06 Nm backdrive torque, 10.15$\pm$1.34 Hz torque bandwidth, $\pm 40^\circ$ ROM in both DOFs, and a diagonal velocity-constraint Jacobian (one-to-one motor-to-DOF mapping) in a 0.97 kg package. In practice, these properties increase workspace in clutter and stabilize contact without finely tuned admittance control. We evaluate DexWrist as a drop-in upgrade in simulation and on three robot arms across constrained and contact-rich tasks. In learned policy evaluations on the AgileX PiPER and UR3e, DexWrist achieved 50-76% relative improvements in success rate and reduced autonomous task completion times by 3-5x; on a torque-controlled Franka FR3, where a strong joint-impedance baseline already succeeds, it still completed the task 1.4x faster. Project page and videos: https://martinpeticco.com/dexwrist

cs.RO↗
arXiv2026

Prompt-Driven Exploration: Language as an Exploration Space for VLA Reinforcement Learning

Exploration is essential to RL since a policy cannot improve by repeatedly sampling the behaviors it already prefers. Standard methods inject stochasticity in the action space, but such jitter only yields rollouts close to the original. Escaping a weak policy often requires global perturbations that action noise cannot produce. Large language models (LLMs) and vision-language-action (VLA) models offer a pathway: they condition the policy on a natural language prompt, and since the rollout follows from it, modifying the prompt induces global changes. The challenge is finding prompts that induce useful global changes. With a weak policy that rarely succeeds, reward is too sparse to select on. Our idea is to refine prompts from the rollouts themselves: a vision-language model (VLM) reasons over the rollout video, diagnoses how the policy responded, and rewrites the prompt to elicit better behavior next time. This procedure resembles posterior sampling, a classical RL exploration framework, at the level of prompts: the VLM maintains an implicit distribution over useful prompts and updates it from observed rollouts. We call this strategy Prompt-Driven Exploration (PDE). Across manipulation and reasoning tasks, PDE enables RL to learn successful policies even from zero-reward starts, and improves sample efficiency more broadly. Our website is available at https://xinyunsunshine.github.io/prompt-rl.

cs.LG↗
arXiv2026

KaRMA: A Kinematic Metric for Fine Manipulation Ability in Robotic Hands

Traditional robotic hand metrics focus on static properties such as workspace, manipulability, and grasp stability. However, these metrics do not directly measure dexterity under the standard definition in robotic manipulation: the ability to continuously change an object's pose within the hand while maintaining contact from an initial grasp. We introduce Kinematic Rolling Manipulation Ability (KaRMA), a kinematic-only metric for fine manipulation that quantifies reachable in-hand translation and reorientation of a spherical test object within a two-finger precision pinch through feasible rolling motions. KaRMA enforces joint limits, collision constraints, rolling contact, and antipodal force feasibility, then explores reachable in-hand object poses via breadth-first search over translation and rotation primitives. KaRMA reports three scores: translational coverage (KaRMA-T), rotational coverage (KaRMA-R), and sensitivity to the initial grasp (KaRMA-S). We evaluate KaRMA on 16 widely used robotic hands and compare against static baselines, showing that KaRMA separates hands that rank identically under static proxies, reveals translation-rotation tradeoffs invisible to existing baselines, and is qualitatively consistent with selected published task benchmarks where Jacobian-based metrics can be misleading. Code and interactive demos are available at https://martinpeticco.com/karma.

cs.RO↗
arXiv2026

H2LooP Telecom Model v1: From Telecom Comprehension to Autonomous Issue and PR Resolution

We present H2LooP Telecom Model v1, a domain-specialized large language models fine-tuned for the telecommunications industry. We release two domain-adapted model variants serving complementary use cases: a comprehension-focused variant for telecom domain question answering and reasoning, and an agentic variant for autonomous telecom code generation, pull request resolution, and code commits on production repositories. H2LooP Telecom achieves strong results on the GSMA Open Telecom Lite (OT-Lite) benchmark and a proprietary telecom code generation benchmark, outperforming frontier closed-source models such as GPT-5 and Claude Opus on independent leaderboard evaluation, while preserving general-purpose capabilities. The Comprehension variant achieves 81.8% weighted average on OT-Lite Pass@3, and, independently, ranks 5th overall on the official community-run Open Telco AI Leaderboard* at only 31B parameters-ahead of frontier closed-source systems including Claude Opus 4.6, GPT-5, Gemini 3 Flash, Grok-4-fast, and Kimi K2.5. Our agentic variant obtains a relative improvement of +8.8% in AST Similarity and +20.0% in Location IoU over the base model on telecom code generation, while maintaining identical MMLU (74.0%) and BFCL v3 multi-turn function calling (79.0%) performance, indicating zero catastrophic forgetting. Domain specialization on curated telecom corpora, spanning 3GPP standards, O-RAN specifications, network telemetry, and real repository commits, yields substantial improvements over general-purpose models of equivalent scale, approaches frontier closed-source models on domain-specific evaluation, and is independently corroborated by our official leaderboard standing.

cs.CL↗
arXiv2026

Dynamic Compression in Recurrent Networks

Recurrent models process long contexts efficiently by compressing their history into a fixed-size state, but modern architectures typically do so in a single causal pass over the sequence. Each input must therefore be compressed before the model knows how it will later be used, forcing a limited state to compromise across possible future demands. We introduce dynamic compression, which allows a recurrent model to selectively revisit past tokens and revise its fixed-size state through additional recurrent updates. The model need not preserve every part of the history at uniformly high fidelity in its recurrent state, because lower-fidelity information can be revisited from the retained raw sequence when it becomes relevant. We study this in a controlled setting where the model first learns multiple functions in-context and, later in the same sequence, encounters a series of few-shot tasks that each require it to identify and reuse one of those functions. A single-pass model must preserve every function at sufficient fidelity for any future task, whereas selective re-scanning allows the model to revisit and refine only the function currently needed. We find that dynamic compression substantially reduces the recurrent state required for accurate reuse and scales more favorably as the number of stored functions grows. These results demonstrate a computation--memory tradeoff in which recurrent models can spend more computation revisiting their history to make more effective use of a fixed-size state.

cs.LG↗
arXiv2026

Self-Distillation Enables Continual Learning

Continual learning, enabling models to acquire new skills and knowledge without degrading existing capabilities, remains a fundamental challenge for foundation models. While on-policy reinforcement learning can reduce forgetting, it requires explicit reward functions that are often unavailable. Learning from expert demonstrations, the primary alternative, is dominated by supervised fine-tuning (SFT), which is inherently off-policy. We introduce Self-Distillation Fine-Tuning (SDFT), a simple method that enables on-policy learning directly from demonstrations. SDFT leverages in-context learning by using a demonstration-conditioned model as its own teacher, generating on-policy training signals that preserve prior capabilities while acquiring new skills. Across skill learning and knowledge acquisition tasks, SDFT consistently outperforms SFT, achieving higher new-task accuracy while substantially reducing catastrophic forgetting. In sequential learning experiments, SDFT enables a single model to accumulate multiple skills over time without performance regression, establishing on-policy distillation as a practical path to continual learning from demonstrations.

cs.LG↗
arXiv2026

Fine-Tuning and Serving Gemma 4 31B on Google Cloud TPU: A Technical Comparison with GPU Baselines

We present the first end-to-end demonstration of fine-tuning and serving Google's Gemma 4 31B model on TPU hardware, providing an empirical comparison of TPU and GPU platforms for large language model adaptation. Using LoRA on a Google TPU v5p-8 for training and TPU v6e-8 (Trillium) for inference, we document the full set of code-level adaptations required to port a GPU-native training recipe - built on PyTorch, HuggingFace TRL, and FSDP - to the JAX + Tunix/Qwix stack. These adaptations span mesh configuration, LoRA module naming conventions, sharding annotation corrections, gradient checkpoint, data pipeline restructuring, and a custom Orbax-to-safetensor checkpoint merging procedure. For inference, we detail the vLLM-TPU Docker setup necessary to serve Gemma 4 on v6e-8 and characterize the resulting latency and throughput profile. Compared with a similar-costing 2xH100 GPU baseline under identical hyperparameters, TPU training completes 1.61x faster at 2.12x lower cost. For inference, we cover the vLLM-TPU Docker setup required to serve Gemma 4 on v6e-8 and explain the observed latency and throughput characteristics across a QPS sweep spanning 512 to 16k input tokens. Across both workloads we compare performance and cost against a 2xH100 GPU baseline running identical hyperparameters. The TPU completes training 1.61x faster at 2.12x lower cost. For inference, TPU v6e-8 matches GPU at short context (<=2048 tokens) and decisively outperforms at long context: 66% higher throughput and 23.6x faster TTFT at 4096-token inputs (61 ms vs 1,443 ms at QPS=4). Our work removes a critical gap in the open tooling ecosystem and provides practitioners with a recipe for Gemma 4 Dense 31B deployment on the TPU infrastructure.

cs.DC↗
arXiv2026

What Matters When Cotraining Robot Manipulation Policies on Everyday Human Videos?

Human video datasets used for cotraining robot manipulation policies largely consist of curated demonstrations where motions are orchestrated to resemble robot behavior and 3D hand poses are captured with specialized hardware. A more plentiful source of data is everyday Internet video, but it is an open question what factors enable transfer from such videos to robots. We investigate this using a new dataset of 532 human videos with 28 hours of high-quality triangulated hand labels and natural motions. We find that hand pose quality affects transfer, but even with accurate hands, the inherent motion gap hinders transfer unless the vision and policy networks specialize to each embodiment. Our cotraining recipe yields consistent improvements, with an absolute success rate gain of $29.7\%$ in the low-robot-data regime across six manipulation tasks.

cs.RO↗
arXiv2026

Vector Policy Optimization: Training for Diversity Improves Test-Time Search

Language models must now generalize out of the box to novel environments and work inside inference-scaling search procedures, such as AlphaEvolve, that select rollouts with a variety of task-specific reward functions. Unfortunately, the standard paradigm of LLM post-training optimizes a pre-specified scalar reward, often leading current LLMs to produce low-entropy response distributions and thus to struggle at displaying the diversity that inference-time search will require. We propose Vector Policy Optimization (VPO), an RL algorithm that explicitly trains policies to anticipate diverse downstream reward functions and to produce diverse solutions. VPO exploits that rewards are often vector-valued in practice, like per-test-case correctness in code generation or, say, multiple different user personas or reward models. VPO is essentially a drop-in replacement for the GRPO advantage estimator, but it trains the LLM to output a set of solutions where individual solutions specialize to different trade-offs in the vector reward space. Across four tasks, VPO matches or beats the strongest scalar RL baselines on test-time search (e.g. pass@k and best@k), with the gap widening as the search budget grows. For evolutionary search, VPO models unlock problems that GRPO models cannot solve at all. As test-time search becomes more standardized, optimizing for diversity may need to become the default post-training objective.

cs.LG↗
arXiv2026

Tune to Learn: How Controller Gains Shape Robot Policy Learning

Position controllers have become the dominant interface for executing learned manipulation policies. Yet a critical design decision remains understudied: how should we choose controller gains for policy learning? The conventional wisdom is to select gains based on desired task compliance or stiffness. However, this logic breaks down when controllers are paired with state-conditioned policies: effective stiffness emerges from the interplay between learned reactions and control dynamics, not from gains alone. We argue that gain selection should instead be guided by learnability: how amenable different gain settings are to the learning algorithm in use. In this work, we systematically investigate how position controller gains affect three core components of modern robot learning pipelines: behavior cloning, reinforcement learning from scratch, and sim-to-real transfer. Through extensive experiments across multiple tasks and robot embodiments, we find that: (1) behavior cloning benefits from compliant and overdamped gain regimes, (2) reinforcement learning can succeed across all gain regimes given compatible hyperparameter tuning, and (3) sim-to-real transfer is harmed by stiff and overdamped gain regimes. These findings reveal that optimal gain selection depends not on the desired task behavior, but on the learning paradigm employed. Project website: https://younghyopark.me/tune-to-learn

cs.RO↗