3arXiv cs.AI (Artificial Intelligence)·
SGA: Plug-and-play geometric verification module for LLM-generated educational animations
Researchers introduce the Symbolic Geometric Agent (SGA), a modular component that intercepts LLM-generated Manim animation code, extracts symbolic scene graphs via partial execution, and applies targeted refinement to resolve spatial conflicts and occlusions. They also propose MVQS (Manim Visual Quality Score), a deterministic rendering-free proxy metric for spatial integrity. Evaluated on the MMMC-Code benchmark across four LLM backbones and two agentic pipelines, SGA achieves a 16.1% relative improvement in MVQS over baseline. The work addresses a gap in LLM-driven code generation pipelines where geometric correctness is underserved.
Related guides (1)
Related events (8)
CARLA-GS: Modular pipeline for photorealistic autonomous driving corner-case synthesis using LLMs and Gaussian splatting
CARLA-GS is a new modular framework for synthesizing safety-critical corner cases for autonomous driving evaluation, decoupling visual representation (editable Gaussian scenes), semantic reasoning (multi-agent LLM for waypoint generation), and physics simulation (CARLA with PID control). The pipeline reconstructs real driving scenes from the Waymo Open Dataset, uses LLMs to identify risky interactions and generate intent-level trajectories, then re-projects simulated states into Gaussian scenes for photorealistic ego-centric rendering. The approach addresses a key gap in AV safety evaluation: generating spatiotemporally consistent, physically feasible, and semantically meaningful rare scenarios at scale.
SIGA: Self-evolving grounding adapters enable coding agents to operate scientific simulators
SIGA (Simulator-Interface Grounding Adapter) is a lightweight adapter framework that equips general-purpose coding agents with the executable contracts needed to configure and run specialized scientific simulators. Evaluated primarily on GEOS (a multiphysics subsurface simulator), SIGA achieves a ~36x wall-clock speedup over human experts and improves TreeSim scores from 0.720 to 0.789 on held-out tasks, with self-evolution via trajectory rewriting yielding further gains. The system also transfers to OpenFOAM and LAMMPS, revealing that the dominant grounding mechanism (validation vs. memory/retrieval) shifts depending on the interface type. The work frames simulator setup as an agent-tool interface grounding problem, offering a generalizable pattern for deploying coding agents on domain-specific software.
Gazer: Training-free semantic correction for autoregressive visual models using MLLM feedback
Researchers introduce Gazer, a training-free framework that integrates multimodal large language model feedback into the sampling loop of autoregressive visual models (AVMs) to correct semantic errors during generation. The system operates in two stages: Reflective Diagnosis identifies semantic errors in intermediate generation states, and Semantic Correction rewinds and adjusts the generation trajectory to better match the target prompt. Experiments on compositional image and video benchmarks show improved semantic alignment and compositional accuracy across multiple AVMs without additional training. The work addresses a known weakness of next-scale prediction AVMs, where semantic errors accumulate across discrete generation scales.
SD-GPS: Solver-Driven Autoformalization and Theorem Proposing for Geometry Problem Solving
Researchers propose SD-GPS, a neuro-symbolic framework for geometry problem solving that treats a symbolic solver as an execution oracle during both formalization and deduction stages. The system combines solvability-guided reinforcement learning for autoformalization (built on QwenVL3-2B) with an impasse-aware agent that proposes and symbolically verifies auxiliary lemmas. Evaluations on Geometry3K and PGPS9K show SD-GPS outperforms existing multimodal, neural, and neuro-symbolic baselines across multiple task regimes. The work advances the line of research on grounding neural agents in formal systems for verifiable mathematical reasoning.
Geometric Action Model (GAM) repurposes geometric foundation models for 3D-aware robot manipulation
Researchers propose the Geometric Action Model (GAM), a language-conditioned robot manipulation policy that splits a pretrained geometric foundation model (GFM) to serve simultaneously as an observation encoder, causal future predictor, and action decoder. Unlike existing vision-language-action models that operate on 2D image frames, GAM explicitly incorporates 3D geometric priors for contact-rich manipulation. The approach claims improvements in accuracy, robustness, speed, and model size over foundation-model-scale baselines across simulation and real-robot benchmarks.
FormalAnalyticGeo: Neural-symbolic framework for automatic analytic geometry problem generation yields 7K-problem dataset
Researchers introduce FormalAnalyticGeo, a pipeline using four specialized LLM components and a formal intermediate language (CDL) to automatically generate multimodal analytic geometry problems with diagram rendering via a Signed Distance Field engine. The closed-loop system requires no human annotation and produces AnalyticGeo7K, a dataset of over 7,000 verified problems with aligned text, diagrams, formal annotations, and ground-truth answers. Generated problems achieve a median ground-truth relative error of 0.70%, with 82.3% of answers within 5% of exact symbolic solutions. The work addresses a recognized gap in math reasoning benchmarks where analytic geometry is underrepresented due to data scarcity.
LLM-as-a-Verifier: Training-free verification framework scales along granularity, repetition, and criteria decomposition
Researchers introduce LLM-as-a-Verifier, a general-purpose verification framework that treats verification as a new scaling axis for LLMs, computing continuous scores from token logit distributions rather than discrete judge outputs. The framework scales along three dimensions—score granularity, repeated evaluation, and criteria decomposition—and achieves state-of-the-art results on Terminal-Bench V2 (86.5%), SWE-Bench Verified (78.2%), RoboRewardBench (87.4%), and MedAgentBench (73.3%) without requiring additional training. The authors also demonstrate that the framework's fine-grained signals can serve as dense RL feedback, improving sample efficiency for SAC and GRPO on robotics and math benchmarks, and build a Claude Code extension for monitoring agentic systems.
Symbal: Automated detection of systematic misalignments in MLLM-generated image captions
Researchers from Stanford introduce Symbal, a dual-stage framework using off-the-shelf foundation models to detect recurring, visually-correlated errors in multimodal LLM-generated captions. They also release SymbalBench, a benchmark of 1.7 million image-text pairs across 420 vision-language datasets covering natural and medical imaging domains. Symbal correctly identifies systematic misalignments in 63.8% of datasets, nearly 4x better than the closest baseline, and works without access to the underlying MLLM. The work addresses a practical data-quality auditing problem relevant to anyone building or evaluating vision-language datasets.