Skip to main content
Search

New York University

34 related articles

Subscribe to updates
  1. arXiv

    Agents no longer learn from a scalar reward alone: ACE-RL trains the reflections and notes they hand forward, topping held-out reward on four benchmarks

    Agent reinforcement learning usually compresses a trial into one scalar reward, discarding the error logs and tool outputs the environment revealed; ACE-RL also trains the reflections and task notes the model writes, scoring each candidate note by a fresh attempt that executes it. At equal training compute on Mastermind, MBPP, AppWorld and MiniSWEBench, ACE-RL's held-out reward is higher than actor-only GRPO, ERL, and delayed-joint PPO trained end to end through the harness with a learned critic.

  2. arXiv

    Neural networks find high-amplitude water-wave candidates, and interval arithmetic plus Lean turn the residual bound into a checkable statement

    Physics-informed neural networks could previously show only small residuals at sampled points, not how large the error is over the continuous domain. This pipeline has a network search the Babenko equation in conformal coordinates for high-amplitude gravity–capillary waves, and staged neural corrections, Fourier projection and Newton polishing cut the maximum sampled residual by about seven orders of magnitude. The candidate is then frozen as an explicit trigonometric polynomial, interval arithmetic yields a rigorous uniform bound on the continuous residual, and Lean machine-checks the deductions from that numerical bound to the final residual theorem.

  3. arXiv

    Robot-arm inverse kinematics moves to a learned closed loop: PolicyIK holds 1 mm on 93.5% of held-out tasks in one attempt

    Robot-arm inverse kinematics is usually handled by a numerical solver plus a hand-tuned controller; PolicyIK replaces that stage with a goal-conditioned reinforcement-learning policy that updates joint references from pose feedback through a joint-position interface. On a simulated 7-DoF OpenArm, the trained PPO policy keeps the mean error over the final second within 1 mm on 93.5% of held-out tasks in a single attempt, and a deployment protocol raises success to 98.9%; trained directly on a physical arm, it reaches and holds more tasks than four classical IK controllers.

  4. Robots stop taking commands literally: MIT's FLEX uses an LLM to turn "don't spill on the laptop" into avoiding every device

    Researchers at MIT and New York University introduce FLEX, a flexible feature-editing pipeline for robot motion planning that lets a robotic arm go beyond the literal wording of a natural-language instruction and infer a higher-level intent the person never stated. Earlier methods added one object to an avoidance list per correction, so the list grew long and went stale as the environment changed. FLEX uses a large language model to generalize across the features already named and a vision system to drop objects no longer present; in three simulated tasks, serving coffee, moving a candle and carrying a bowling ball, the robot followed the literal command and extended avoidance to electronics, flammable objects and fragile items.

  5. arXiv

    Phylogenetic posteriors get an iterative generative model: it matches IQ-Tree on simple data and beats it on complex evolution

    Posterior inference over phylogenetic trees has depended on likelihoods, which become intractable once real processes such as insertions, deletions or site selection are modeled. This family of conditional generative models builds a tree step by step, bottom-up, middle-out or top-down, and is trained with a discrete diffusion objective; on real and synthetic data it recovers MCMC posteriors more closely than the prior simulation-based method Phyloformer 2, matches IQ-Tree on simple data, and is reliably more accurate on more realistic complex models.

  6. arXiv

    Humanoid robots learn how hard to press: FLoRe captures contact force from robot-free demos and reaches 80% success on three contact-rich tasks

    Everyday humanoid work such as pushing a cart, moving a box or wiping a surface needs controlled contact force, not just tracked hand motion. FLoRe makes force a shared interface across demonstration, policy and whole-body control: two handheld grippers with six-axis force/torque sensors record hand motion and contact wrenches without a robot, a diffusion policy predicts per-hand force commands from those demonstrations, and a whole-body controller tracks them by integrating the measured force error. On a Unitree G1 across three contact-rich tasks, the force-reactive policy succeeded in 80% of rollouts, against 26.7% for a position-only policy trained on the same demonstrations and run by the same controller.

  7. arXiv

    Training logs know what deployment cannot: runtime-confounded RL gets doubly robust evaluation and a testable price

    When deployment lacks key variables that training logs record in full, runtime confounding can push a reinforcement-learning policy off course over time. The work places the runtime restriction only in the numerator of an importance ratio and estimates nuisance functions from the full data, yielding an asymptotically normal off-policy estimator that attains the semiparametric efficiency bound, a policy-learning regret bound that differs from the full-information optimum by an irreducible runtime gap, and a sample-splitting test for that gap.

  8. arXiv

    Alignment without scale: good views let a 109M text JEPA beat 72B language models on the Platonic representation benchmark

    The Platonic Representation Hypothesis, which holds that separately trained models converge on a shared picture of the world, becomes a theorem for JEPAs (joint-embedding predictive architectures): if each modality's positive pairs link the same world states, the encoders agree up to one rotation. The resulting good-view versus bad-view distinction predicts alignment, and a 109M-parameter text JEPA reaches .184 on the PRH benchmark, above every language model evaluated, up to Qwen2.5-72B at .154.

  9. arXiv

    Multi-agent workflows stop chasing task score alone: LAWA edits the critical path and cuts makespan on seven benchmarks

    Automatic multi-agent workflow generation now targets execution time: LAWA cuts total makespan through graph edits aimed at critical-path bottlenecks while keeping competitive task performance across seven benchmarks.

  10. arXiv

    Corrupted distances, still a near-optimal spanning tree: logarithmic approximation under arbitrary tampering, with no true-distance queries

    In a metric whose pairwise distances are each corrupted with fixed probability by an adversary, an algorithm sees only the corrupted values. For increase-only, decrease-only and unrestricted corruption, the work gives polynomial-time or quasi-polynomial-time approximations, reaching a logarithmic factor in the unrestricted case that matches the known information-theoretic lower bound, all without querying true distances.