# LLM agent

When a mobile manipulator is told to put the toys away in the toy box, it usually relies on a task-agnostic semantic map that may lack the attributes, affordances or concepts the instruction needs. Feijoa has a language-model agent query a posed RGB-D scan after the instruction arrives, inspect the returned candidates and rewrite its queries until the reference resolves. On 103 open-ended grounding queries it reaches 94.2% exact match, against 68.0% for the best prebuilt representation, TextMap; on 50 real Spot trials it completes 35 tasks, 17 of them after recovering from a failed tool call or verification check.

When a mobile manipulator is told to put the toys away in the toy box, it usually relies on a task-agnostic semantic map that may lack the attributes, affordances or concepts the instruction needs. Feijoa has a language-model agent query a posed RGB-D scan after the instruction arrives, inspect the returned candidates and rewrite its queries until the reference resolves. On 103 open-ended grounding queries it reaches 94.2% exact match, against 68.0% for the best prebuilt representation, TextMap; on 50 real Spot trials it completes 35 tasks, 17 of them after recovering from a failed tool call or verification check.

Physical hardware security evaluation has depended on a person choosing each laser fault-injection position, power and timing. A large-language-model agent now links firmware reverse engineering, chip imagery and experimental decisions into a closed loop, choosing the next shot inside human-approved safety bounds. On an ATmega328P it produced a skipped load instruction and a wrong-PIN acceptance confirmed by a separate state read; on an Artix-7 FPGA the same workflow extended to ciphertext-error classification in programmable logic.

Physical hardware security evaluation has depended on a person choosing each laser fault-injection position, power and timing. A large-language-model agent now links firmware reverse engineering, chip imagery and experimental decisions into a closed loop, choosing the next shot inside human-approved safety bounds. On an ATmega328P it produced a skipped load instruction and a wrong-PIN acceptance confirmed by a separate state read; on an Artix-7 FPGA the same workflow extended to ciphertext-error classification in programmable logic.

A home-care scheduling agent no longer asks a small model to handle a narrow task cold. Zero-shot small models failed it, and retraining on every behavior change was too slow, so the behavior was first iterated in prompt space on GPT-5.4-nano, then 11.6k recorded production decisions were LoRA-tuned into Qwen3-8B. That model agrees with recorded decisions 0.913 of the time, above the teacher's 72.7% self-agreement when re-run on its own prompts.

A home-care scheduling agent no longer asks a small model to handle a narrow task cold. Zero-shot small models failed it, and retraining on every behavior change was too slow, so the behavior was first iterated in prompt space on GPT-5.4-nano, then 11.6k recorded production decisions were LoRA-tuned into Qwen3-8B. That model agrees with recorded decisions 0.913 of the time, above the teacher's 72.7% self-agreement when re-run on its own prompts.

JWST/NIRCam F444W slitless spectroscopy over 0.33 square degrees in COSMOS has yielded a homogeneous catalog of 12,275 emission-line galaxies spanning redshift 0 to 9, now publicly released.

JWST/NIRCam F444W slitless spectroscopy over 0.33 square degrees in COSMOS has yielded a homogeneous catalog of 12,275 emission-line galaxies spanning redshift 0 to 9, now publicly released.

DRYLAB-Bench puts 15 frontier models through 30 hazardous mutation-design scenarios and scores submitted amino-acid substitutions against published wet-lab measurements: 21.4% land in each assay's highest-risk decile without tools, rising to 23.5% with biological design tools and multi-round iteration.

DRYLAB-Bench puts 15 frontier models through 30 hazardous mutation-design scenarios and scores submitted amino-acid substitutions against published wet-lab measurements: 21.4% land in each assay's highest-risk decile without tools, rising to 23.5% with biological design tools and multi-round iteration.

The authors present QuSema, an autonomous testing agent that uses quantum semantics and documentation as a source-level semantic oracle, running a three-stage loop of contextual unit preparation, semantic defect detection, and library API triggering and validation to decide whether implementation logic can turn valid inputs into invalid outputs; on a benchmark of 20 historical silent bugs in Qiskit and PennyLane it achieves higher mean bug relocation counts than Claude Code and Codex, and it discovers 40 previously unknown bugs confirmed by maintainers, 30 of them silent.

The authors present QuSema, an autonomous testing agent that uses quantum semantics and documentation as a source-level semantic oracle, running a three-stage loop of contextual unit preparation, semantic defect detection, and library API triggering and validation to decide whether implementation logic can turn valid inputs into invalid outputs; on a benchmark of 20 historical silent bugs in Qiskit and PennyLane it achieves higher mean bug relocation counts than Claude Code and Codex, and it discovers 40 previously unknown bugs confirmed by maintainers, 30 of them silent.

The work introduces Task-Progress Distillation (TPD), an offline approach that pairs each demonstrated action with a short task-stage label so the student selects actions by jointly scoring admissible stage–action pairs; on ALFWorld, a 1.7B student trained with 404 demonstrations reaches 72.4% mean unseen task success with either TPD or action-only supervision, versus 48.3% for a reasoning-trained student using constrained action selection, and explicit stages raise success from 48.0% to 67.7% at 200 demonstrations, with action-only catching up to 76.9% at 808.

The work introduces Task-Progress Distillation (TPD), an offline approach that pairs each demonstrated action with a short task-stage label so the student selects actions by jointly scoring admissible stage–action pairs; on ALFWorld, a 1.7B student trained with 404 demonstrations reaches 72.4% mean unseen task success with either TPD or action-only supervision, versus 48.3% for a reasoning-trained student using constrained action selection, and explicit stages raise success from 48.0% to 67.7% at 200 demonstrations, with action-only catching up to 76.9% at 808.

The work introduces PoS, an inference-time framework that constructs and continually maintains explicit belief states as the agent's decision context, where each belief combines an estimate of the current world state with unresolved task requirements; PoS validates belief consistency and monitors task progress to detect Belief Trapping, then tailors recovery to the trapping pattern and the type of unresolved requirement. On four benchmarks spanning execution and diagnosis, PoS achieves the highest overall performance with all three LLM backbones; ablations show the importance of consistency validation and recovery, and context-scaling experiments show resilience to context growth.

The work introduces PoS, an inference-time framework that constructs and continually maintains explicit belief states as the agent's decision context, where each belief combines an estimate of the current world state with unresolved task requirements; PoS validates belief consistency and monitors task progress to detect Belief Trapping, then tailors recovery to the trapping pattern and the type of unresolved requirement. On four benchmarks spanning execution and diagnosis, PoS achieves the highest overall performance with all three LLM backbones; ablations show the importance of consistency validation and recovery, and context-scaling experiments show resilience to context growth.