A compiler for a self-hosting Scheme subset produces a differentiable meta-circular interpreter supporting gradient-based co-optimization of program structure and continuous parameters.
Canonical reference
Pawan Kumar, Emilien Dupont, Francisco J
Canonical reference. 100% of citing Pith papers cite this work as background.
60
Pith papers citing it
315
external citations · Crossref
Background
100% of classified citations
citation-role summary
background 9
citation-polarity summary
fields
cs.AI 27 cs.LG 11 cs.CL 6 cs.CE 2 cs.DC 2 cs.NE 2 quant-ph 2 cond-mat.mtrl-sci 1 cs.PL 1 cs.RO 1roles
background 9polarities
background 9representative citing papers
Frontier coding agents surpass published Nature-family SOTA on only 17.8% of 90 sealed scientific tasks, mostly by recasting problems as supervised ML rather than inventing methods.
An LLM-driven agent with built-in seed-noise audits develops control policies for two aerospace problems that outperform undirected search and pass verification checks.
Proves R(B_8, B_10) = 37 via an AI-assisted short proof with a Lean formalization of the upper bound.
GAMBLe decomposes ADRS into four parameters and an effective landscape, with experiments on 760+ runs across NP-hard problems showing no universal best generator or mechanism and potential gains of 13-67% from component choice.
Evolutionary coding agents achieve most benchmark gains through a small subset of edit types and by cycling previously deleted code lines rather than developing new algorithmic structures.
PopPy combines an ahead-of-time compiler and runtime to extract parallelism from Python compound AI applications, delivering up to 6.4x end-to-end speedups while preserving sequential semantics.
Latent Heuristic Search performs continuous optimization over learned embeddings of heuristics, using normalizing flows and LLM prompting to discover competitive solvers for TSP, CVRP, KSP, and OBP.
A finite sheaf-theoretic framework ranks obstruction measures to identify when an AI agent's theory must deform within its language or extend to a new one, validated on a controlled transition benchmark.
An interactive AI workbench for mathematicians achieves 48% on FrontierMath Tier 4 and helped solve open problems in early tests.
LOVER creates an unsupervised logic-regularized verifier that reaches 95% of supervised verifier performance on reasoning tasks across 10 datasets.
AI coding agents evolve simple ground-state protocols into improved versions for VQE, DMRG, and AFQMC on spin models and molecules by using executable energy scores under fixed compute budgets.
Vision-language models underperform specialized astronomical methods on real observational data, with accuracy improving when physical explanations are provided in prompts and when raw numerical measurements replace rendered plots.
EVIL evolves single compact Python algorithms via LLM-guided search that enable zero-shot inference on temporal point processes, Markov jump processes, and time series imputation, often matching or exceeding deep learning models while remaining interpretable and orders of magnitude faster.
SignalClaw synthesizes interpretable, composable traffic signal control skills through LLM-guided evolution that matches top baselines on routine SUMO scenarios and outperforms them on emergency and transit events while remaining editable by engineers.
An agentic AI framework using latent models as fast surrogates autonomously explores PDE parameter spaces and identifies regime-dependent scaling laws in tandem cylinder wake flows.
AlphaEvolve is an LLM-orchestrated evolutionary coding agent that discovered a 4x4 complex matrix multiplication algorithm using 48 scalar multiplications, the first improvement over Strassen's algorithm in 56 years, plus optimizations for Google data centers and hardware.
Introduces an auditable four-stage diagnostic for LLM physics reasoning in novel frameworks and applies it to three parallel worlds, yielding pass rates of 6/15, 6/15, and 0/15 on frontier models with noted qualitative-quantitative asymmetry.
Transformers generate new FRSTs of 4D reflexive polytopes across size ranges and self-improve by retraining on their own outputs.
Geometry-aware MCTS with incremental constraint updates and symmetry pruning yields new best-known configurations for five of six tested combinatorial geometry problems, including ~1.8n points for Max-N3IL on grids 82-119.
Heuresis evaluates six search strategies for autonomous ML research agents and finds that novel ideas are rare, none rated original, and only one reaches top-10 quality while strategies steer axes but do not expand the quality-novelty frontier.
RACL lets a reasoning agent discover and apply control rules to a metaheuristic by observing operational memory and testing bounded interventions, shown on vehicle routing with reported cost improvements over baselines.
Evoflux applies evolutionary search at inference time to repair executable tool workflows for compact agents, outperforming SFT and SFT+DPO on held-out MCP-Bench tasks with live servers and 250 tools.
Arbor applies tree search as a cognition layer across Orchestrator and Critic agents to run multi-day autonomous optimization campaigns, reporting up to 193% throughput-latency gains over vendor baselines on LLM inference.
citing papers explorer
-
Compile Once, Differentiate Everywhere: A Differentiable Meta-Circular Interpreter
A compiler for a self-hosting Scheme subset produces a differentiable meta-circular interpreter supporting gradient-based co-optimization of program structure and continuous parameters.
-
NatureBench: Can Coding Agents Match the Published SOTA of Nature-Family Papers?
Frontier coding agents surpass published Nature-family SOTA on only 17.8% of 90 sealed scientific tasks, mostly by recasting problems as supervised ML rather than inventing methods.
-
Agentic AutoResearch forSpace Autonomy: An Auditable, LLM-Driven Research Agent for Aerospace Control Problems
An LLM-driven agent with built-in seed-noise audits develops control policies for two aerospace problems that outperform undirected search and pass verification checks.
-
An automated proof that R(B_8,B_10)=37
Proves R(B_8, B_10) = 37 via an AI-assisted short proof with a Lean formalization of the upper bound.
-
Don't Gamble, GAMBLe: An Analytical Framework for AI-Driven Research Systems
GAMBLe decomposes ADRS into four parameters and an effective landscape, with experiments on 760+ runs across NP-hard problems showing no universal best generator or mechanism and potential gains of 13-67% from component choice.
-
What Do Evolutionary Coding Agents Evolve?
Evolutionary coding agents achieve most benchmark gains through a small subset of edit types and by cycling previously deleted code lines rather than developing new algorithmic structures.
-
PopPy: Opportunistically Exploiting Parallelism in Python Compound AI Applications
PopPy combines an ahead-of-time compiler and runtime to extract parallelism from Python compound AI applications, delivering up to 6.4x end-to-end speedups while preserving sequential semantics.
-
Latent Heuristic Search: Continuous Optimization for Automated Algorithm Design
Latent Heuristic Search performs continuous optimization over learned embeddings of heuristics, using normalizing flows and LLM prompting to discover competitive solvers for TSP, CVRP, KSP, and OBP.
-
Sheaf-Theoretic Transport and Obstruction for Detecting Scientific Theory Shift in AI Agents
A finite sheaf-theoretic framework ranks obstruction measures to identify when an AI agent's theory must deform within its language or extend to a new one, validated on a controlled transition benchmark.
-
AI co-mathematician: Accelerating mathematicians with agentic AI
An interactive AI workbench for mathematicians achieves 48% on FrontierMath Tier 4 and helped solve open problems in early tests.
-
Logic-Regularized Verifier Elicits Reasoning from LLMs
LOVER creates an unsupervised logic-regularized verifier that reaches 95% of supervised verifier performance on reasoning tasks across 10 datasets.
-
Optimizing ground state preparation protocols with autoresearch
AI coding agents evolve simple ground-state protocols into improved versions for VQE, DMRG, and AFQMC on spin models and molecules by using executable energy scores under fixed compute budgets.
-
A systematic evaluation of vision-language models for observational astronomical reasoning tasks
Vision-language models underperform specialized astronomical methods on real observational data, with accuracy improving when physical explanations are provided in prompts and when raw numerical measurements replace rendered plots.
-
EVIL: Evolving Interpretable Algorithms for Zero-Shot Inference on Event Sequences and Time Series with LLMs
EVIL evolves single compact Python algorithms via LLM-guided search that enable zero-shot inference on temporal point processes, Markov jump processes, and time series imputation, often matching or exceeding deep learning models while remaining interpretable and orders of magnitude faster.
-
SignalClaw: LLM-Guided Evolutionary Synthesis of Interpretable Traffic Signal Control Skills
SignalClaw synthesizes interpretable, composable traffic signal control skills through LLM-guided evolution that matches top baselines on routine SUMO scenarios and outperforms them on emergency and transit events while remaining editable by engineers.
-
Agentic Exploration of PDE Spaces using Latent Foundation Models for Parameterized Simulations
An agentic AI framework using latent models as fast surrogates autonomously explores PDE parameter spaces and identifies regime-dependent scaling laws in tandem cylinder wake flows.
-
AlphaEvolve: A coding agent for scientific and algorithmic discovery
AlphaEvolve is an LLM-orchestrated evolutionary coding agent that discovered a 4x4 complex matrix multiplication algorithm using 48 scalar multiplications, the first improvement over Strassen's algorithm in 56 years, plus optimizations for Google data centers and hardware.
-
Testing Frontier Large Language Models' Physics Literacy in Parallel Physical Worlds
Introduces an auditable four-stage diagnostic for LLM physics reasoning in novel frameworks and applies it to three parallel worlds, yielding pass rates of 6/15, 6/15, and 0/15 on frontier models with noted qualitative-quantitative asymmetry.
-
Generating Special Triangulations with Transformers
Transformers generate new FRSTs of 4D reflexive polytopes across size ranges and self-improve by retraining on their own outputs.
-
Geometry-Aware MCTS for Extremal Problems in Combinatorial Geometry
Geometry-aware MCTS with incremental constraint updates and symmetry pruning yields new best-known configurations for five of six tested combinatorial geometry problems, including ~1.8n points for Max-N3IL on grids 82-119.
-
Heuresis: Search Strategies for Autonomous AI Research Agents Across Quality, Diversity and Novelty
Heuresis evaluates six search strategies for autonomous ML research agents and finds that novel ideas are rare, none rated original, and only one reaches top-10 quality while strategies steer axes but do not expand the quality-novelty frontier.
-
RACL: Reasoning-Agent Control Layers for Continuous Metaheuristic Learning
RACL lets a reasoning agent discover and apply control rules to a metaheuristic by observing operational memory and testing bounded interventions, shown on vehicle routing with reported cost improvements over baselines.
-
Evoflux: Inference-Time Evolution of Executable Tool Workflows for Compact Agents
Evoflux applies evolutionary search at inference time to repair executable tool workflows for compact agents, outperforming SFT and SFT+DPO on held-out MCP-Bench tasks with live servers and 250 tools.
-
Arbor: Tree Search as a Cognition Layer for Autonomous Agents
Arbor applies tree search as a cognition layer across Orchestrator and Critic agents to run multi-day autonomous optimization campaigns, reporting up to 193% throughput-latency gains over vendor baselines on LLM inference.
-
Toward Generalist Autonomous Research via Hypothesis-Tree Refinement
Arbor combines a coordinator, executors, and a hypothesis tree to enable cumulative autonomous research, outperforming Codex and Claude Code by over 2.5x on six real tasks and reaching 86.36% Any Medal on MLE-Bench Lite.
-
Graph2Idea:Retrieval-Augmented Scientific Idea Generation with Graph-Structured Contexts
Graph2Idea builds dynamic knowledge graphs from retrieved literature to supply compact, relational contexts that guide LLMs in generating novel, feasible, and high-quality scientific ideas, outperforming flat-text baselines on automatic metrics.
-
Self-Evolving Scientific Agent Discovers Generalizable Physically-Reasoned Fluid Control
An LLM-based self-evolving agent discovers a traveling-wave controller with body-frame guidance and yaw feedback that generalizes to unseen targets for an underactuated fluid swimmer.
-
Mutation Without Variation: Convergence Dynamics in LLM-Driven Program Evolution
LLM-driven program mutation converges to restricted structural attractors, with 87% of chains showing over 93% structural revisits and most variation limited to terminal substitutions, unlike classical GP.
-
LLM-Driven Co-Evolutionary Automated Heuristic Design for Bi-Component Coupled Combinatorial Optimization
CoEvo-AHD is an LLM-driven dual-population co-evolutionary method for automated heuristic design in bi-component coupled combinatorial optimization that achieves competitive results on TTP and TPP.
-
DynaSchedBench: Calibrated Dynamic Scheduling Benchmarks and Observability Paradox in LLM-based Scheduling Agents
DynaSchedBench calibrates DFJSP benchmarks via SESC and SSI, revealing an observability paradox and limited gains from LLM agents over heuristics in dynamic scheduling.
-
DEI: Diversity in Evolutionary Inference for Quality-Diversity Search
DEI shows a heterogeneous four-LLM ensemble achieving 124% higher QD-Score and 28% higher coverage than single-model baselines on Core War at equal compute budget.
-
Toward General Quantum Control with Physics-Informed Large Language Models
VF-QCTRL combines LLMs with physics-informed symbolic reasoning and optimization to produce analytic control protocols that match or exceed conventional solvers across a new 16-task benchmark spanning single/multi-qubit, closed/open, and noisy systems.
-
optimize_anything: A Universal API for Optimizing any Text Parameter
A universal LLM optimizer for text artifacts achieves SOTA results on six tasks including tripling ARC-AGI accuracy and cutting cloud costs by 40% via cross-task transfer and side information.
-
AI-Driven Discovery of Information-Efficient Collider Observables for Interference Measurements
AI-driven symbolic evolution discovers interpretable event-level observables that retain substantially more local Fisher information than angular baselines for CP-sensitive HZ interference in two collider channels.
-
Test-Time Compute for Frozen Embedding Models through Agentic Program Search
Agentic program search over a frozen encoder API yields retrieval programs that improve nDCG@10 on held-out tasks and unseen encoder families with no per-domain training.
-
An Information-Theoretic Criterion for Efficient Data Synthesis
Synthetic data improves models only in information-open generation-training loops with external signals, and coarser signals like binary correctness enable better generalization by converging to the most information-efficient component.
-
Parameter-Efficient Neuroevolution for Diverse LLM Generation: Quality-Diversity Optimization via Prompt Embedding Evolution
QD-LLM applies neuroevolution to prompt embeddings within a quality-diversity framework, producing 46% higher coverage and 41% higher QD-score than QDAIF on HumanEval, MBPP, and creative writing benchmarks.
-
Design Structure Matrix Modularization with Large Language Models
LLM prompting for DSM modularization reaches near-reference quality in 30 iterations on five cases with three models, but domain knowledge impairs performance on complex DSMs, leading to the semantic-alignment hypothesis.
-
Agentic Architect: An Agentic AI Framework for Architecture Design Exploration and Optimization
An LLM-driven agentic system evolves microarchitectural policies for cache replacement, data prefetching, and branch prediction, producing designs that match or exceed prior state-of-the-art in IPC on standard benchmarks.
-
Evaluation-driven Scaling for Scientific Discovery
SimpleTES scales test-time evaluation in LLMs to discover state-of-the-art solutions on 21 scientific problems across six domains, outperforming frontier models and optimization pipelines with examples like 2x faster LASSO and new Erdos constructions.
-
AgentGA: Evolving Code Solutions in Agent-Seed Space
AgentGA optimizes agent seeds with genetic algorithms and parent-archive inheritance to improve autonomous code generation, beating a baseline on 15 of 16 Kaggle competitions.
-
Record-Remix-Replay: Hierarchical GPU Kernel Optimization using Evolutionary Search
R^3 optimizes full scientific applications on GPUs better than tuning kernel parameters or compiler flags alone while running nearly an order of magnitude faster than modern evolutionary search methods.
-
Tracing the Thought of a Grandmaster-level Chess-Playing Transformer
Sparse replacement layers decompose the MLP and attention modules of a chess-playing transformer to reveal verifiable tactical reasoning pathways and parallel computation patterns.
-
Large Language Models as Optimization Controllers: Adaptive Continuation for SIMP Topology Optimization
An LLM acting as real-time controller for SIMP topology optimization parameters outperforms fixed schedules and heuristics, delivering 5.7-18.1% lower compliance on 2D and 3D benchmarks.
-
Differentiable Evolutionary Reinforcement Learning
DERL is a differentiable bi-level method that evolves optimal reward structures for RL policies by composing atomic primitives and using meta-gradients from validation performance.
-
Neuro-Symbolic AI for Analytical Solutions of Differential Equations
SIGS is a neuro-symbolic framework that discovers analytical solutions to PDEs by generating grammar-constrained expressions, embedding them in a topology-regularised latent manifold, and refining structure and coefficients against the PDE residual and boundary/initial conditions.
-
A Single Rewrite Suffices: Empirical Lessons from Production Skill Description Optimization
A single LLM rewrite of skill descriptions using false positive and negative cases matches manual optimization performance in production, with most other pipeline components adding little value.
-
Scientific discovery as meta-optimization: a combinatorial optimization case study
Introduces consensus objective aggregation for meta-optimization of scientific discovery and reports improved scaling and speedup for 3-SAT algorithm discovery using digital MemComputing machines.
-
From Meta Idea to Advanced Mathematical Discovery -- Human-AI Co-Discovery of Sign-Embedding Quantum Algorithms
Human-AI collaboration expanded a meta-idea on rational approximation into sign-embedding quantum algorithms for matrix problems, with humans retaining final judgment on routes and refinements.
-
My Chemical Harness: Evolutionary Molecular Design over Synthetic Pathways with Large Language Model Agents
My Chemical Harness performs evolutionary molecular design by searching over validated synthetic routes with LLMs restricted to high-level preferences, outperforming baselines on an sEH proxy task across multiple metrics.