Agent-BRACE improves LLM agent performance on long-horizon partially observable tasks by 5.3-14.5% through a decoupled belief state of verbalized atomic claims with certainty labels that keeps context length constant.
hub
AlphaEvolve: A coding agent for scientific and algorithmic discovery
80 Pith papers cite this work. Polarity classification is still indexing.
80
Pith papers citing it
abstract
In this white paper, we present AlphaEvolve, an evolutionary coding agent that substantially enhances capabilities of state-of-the-art LLMs on highly challenging tasks such as tackling open scientific problems or optimizing critical pieces of computational infrastructure. AlphaEvolve orchestrates an autonomous pipeline of LLMs, whose task is to improve an algorithm by making direct changes to the code. Using an evolutionary approach, continuously receiving feedback from one or more evaluators, AlphaEvolve iteratively improves the algorithm, potentially leading to new scientific and practical discoveries. We demonstrate the broad applicability of this approach by applying it to a number of important computational problems. When applied to optimizing critical components of large-scale computational stacks at Google, AlphaEvolve developed a more efficient scheduling algorithm for data centers, found a functionally equivalent simplification in the circuit design of hardware accelerators, and accelerated the training of the LLM underpinning AlphaEvolve itself. Furthermore, AlphaEvolve discovered novel, provably correct algorithms that surpass state-of-the-art solutions on a spectrum of problems in mathematics and computer science, significantly expanding the scope of prior automated discovery methods (Romera-Paredes et al., 2023). Notably, AlphaEvolve developed a search algorithm that found a procedure to multiply two $4 \times 4$ complex-valued matrices using $48$ scalar multiplications; offering the first improvement, after 56 years, over Strassen's algorithm in this setting. We believe AlphaEvolve and coding agents like it can have a significant impact in improving solutions of problems across many areas of science and computation.
hub tools
claims ledger
- abstract In this white paper, we present AlphaEvolve, an evolutionary coding agent that substantially enhances capabilities of state-of-the-art LLMs on highly challenging tasks such as tackling open scientific problems or optimizing critical pieces of computational infrastructure. AlphaEvolve orchestrates an autonomous pipeline of LLMs, whose task is to improve an algorithm by making direct changes to the code. Using an evolutionary approach, continuously receiving feedback from one or more evaluators, AlphaEvolve iteratively improves the algorithm, potentially leading to new scientific and practical d
co-cited works
representative citing papers
AutoTTS discovers width-depth test-time scaling controllers through agentic search in a pre-collected trajectory environment, yielding better accuracy-cost tradeoffs than hand-designed baselines on math reasoning tasks at low cost.
VibeServe demonstrates that AI agents can synthesize bespoke LLM serving systems end-to-end, remaining competitive with vLLM in standard settings while outperforming it in six non-standard scenarios involving unusual models, workloads, or hardware.
MappingEvolve applies LLMs through Planner-Evolver-Evaluator agents to evolve technology mapping code, delivering 10.04% area reduction versus ABC and 7.93% versus mockturtle on EPFL benchmarks.
Prism is the first symbolic superoptimizer for tensor programs that uses sGraph for compact representation of program families, two-level search, e-graph equivalence checking, and auto-tuning to achieve up to 2.2x speedup over prior superoptimizers on LLM workloads.
InfiniteScienceGym procedurally generates unbounded scientific repositories with exact ground-truth QA pairs to benchmark LLMs on data reasoning, abstention, and tool use without static datasets.
AEvo introduces a meta-agent that edits the evolution procedure or agent context based on accumulated state, outperforming baselines by 26% relative improvement on agentic benchmarks and achieving SOTA on open-ended tasks.
Pinductor leverages language-model priors to learn POMDP world models from limited trajectories, matching privileged-access methods in performance and exceeding tabular baselines in sample efficiency.
A softmax-weighted centroid of the local top-K documents interpolated with the query improves nDCG@10 for frozen embedding models across seven families on held-out BEIR data.
AssayBench is a new gene-ranking benchmark for phenotypic CRISPR screens that shows zero-shot generalist LLMs outperform both biology-specific LLMs and trainable baselines on adjusted nDCG.
A code-graph and correction-based LLM search framework outperforms full-algorithm generation at equal token budgets on three combinatorial optimization problems.
LLM agents in a solver-aware harness recover global constraints from MIP formulations, generate executable propagation-only handlers for SCIP, and solve five additional MIPLIB 2017 instances.
AHD Agent trains a 4B-parameter LLM via agentic RL to actively use tools for automatic heuristic design, matching or exceeding larger baselines across eight domains with fewer evaluations.
An interactive AI workbench for mathematicians achieves 48% on FrontierMath Tier 4 and helped solve open problems in early tests.
Weblica scales RL training for visual web agents by building thousands of reproducible environments through HTTP caching for stable replays and LLM synthesis from real sites, yielding an 8B model that beats similar open baselines on navigation benchmarks.
A knowledge-first approach to LLM-driven automatic heuristic design in combinatorial optimization yields better discovery efficiency, transfer, and generalization than code-centric baselines by formalizing a distortion-compression trade-off.
New nonlinear formulations for geometric packing problems, solved with FICO Xpress and SCIP, produce improved and first-known solutions for several variants.
Agentic-imodels evolves scikit-learn regressors via an autoresearch loop to jointly boost predictive performance and LLM-simulatability, improving downstream agentic data science tasks by up to 73% on the BLADE benchmark.
Litvak's conjecture on minimizing moments of Gaussian minima is disproved by a cosine-based correlation matrix for small n and p, with a new conjecture proposed that this matrix is the general minimizer, supported conditionally via a volumetric extension of Fejes Tóth's zone conjecture.
LLM adaptive exploration via runtime code execution outperforms static query generation for information extraction from heterogeneous BIM models on the new ifc-bench v2 benchmark.
AHE automates coding-agent harness evolution via component, experience, and decision observability, raising Terminal-Bench 2 pass@1 from 69.7% to 77.0% with transfer gains across models and benchmarks.
LLM agents execute scientific tasks but fail to follow core scientific reasoning norms such as evidence consideration and belief revision based on refutations.
Latent Grammar Flow discovers ODEs by placing grammar-based equation representations in a discrete latent space, using a behavioral loss to cluster similar equations, and sampling via a discrete flow model guided by data fit and constraints.
m(Q_d;4) equals d(d² + 3d + 14)/24 + 1 for infinitely many d, with an O(d)-additive upper bound for all d.
citing papers explorer
-
Agent-BRACE: Decoupling Beliefs from Actions in Long-Horizon Tasks via Verbalized State Uncertainty
Agent-BRACE improves LLM agent performance on long-horizon partially observable tasks by 5.3-14.5% through a decoupled belief state of verbalized atomic claims with certainty labels that keeps context length constant.
-
LLMs Improving LLMs: Agentic Discovery for Test-Time Scaling
AutoTTS discovers width-depth test-time scaling controllers through agentic search in a pre-collected trajectory environment, yielding better accuracy-cost tradeoffs than hand-designed baselines on math reasoning tasks at low cost.
-
VibeServe: Can AI Agents Build Bespoke LLM Serving Systems?
VibeServe demonstrates that AI agents can synthesize bespoke LLM serving systems end-to-end, remaining competitive with vLLM in standard settings while outperforming it in six non-standard scenarios involving unusual models, workloads, or hardware.
-
MappingEvolve: LLM-Driven Code Evolution for Technology Mapping
MappingEvolve applies LLMs through Planner-Evolver-Evaluator agents to evolve technology mapping code, delivering 10.04% area reduction versus ABC and 7.93% versus mockturtle on EPFL benchmarks.
-
Prism: Symbolic Superoptimization of Tensor Programs
Prism is the first symbolic superoptimizer for tensor programs that uses sGraph for compact representation of program families, two-level search, e-graph equivalence checking, and auto-tuning to achieve up to 2.2x speedup over prior superoptimizers on LLM workloads.
-
InfiniteScienceGym: An Unbounded, Procedurally-Generated Benchmark for Scientific Analysis
InfiniteScienceGym procedurally generates unbounded scientific repositories with exact ground-truth QA pairs to benchmark LLMs on data reasoning, abstention, and tool use without static datasets.
-
Harnessing Agentic Evolution
AEvo introduces a meta-agent that edits the evolution procedure or agent context based on accumulated state, outperforming baselines by 26% relative improvement on agentic benchmarks and achieving SOTA on open-ended tasks.
-
Learning POMDP World Models from Observations with Language-Model Priors
Pinductor leverages language-model priors to learn POMDP world models from limited trajectories, matching privileged-access methods in performance and exceeding tabular baselines in sample efficiency.
-
Test-Time Compute for Dense Retrieval: Agentic Program Generation with Frozen Embedding Models
A softmax-weighted centroid of the local top-K documents interpolated with the query improves nDCG@10 for frozen embedding models across seven families on held-out BEIR data.
-
AssayBench: An Assay-Level Virtual Cell Benchmark for LLMs and Agents
AssayBench is a new gene-ranking benchmark for phenotypic CRISPR screens that shows zero-shot generalist LLMs outperform both biology-specific LLMs and trainable baselines on adjusted nDCG.
-
Budget-Efficient Automatic Algorithm Design via Code Graph
A code-graph and correction-based LLM search framework outperforms full-algorithm generation at equal token budgets on three combinatorial optimization problems.
-
Agentic MIP Research: Accelerated Constraint Handler Generation
LLM agents in a solver-aware harness recover global constraints from MIP formulations, generate executable propagation-only handlers for SCIP, and solve five additional MIPLIB 2017 instances.
-
AHD Agent: Agentic Reinforcement Learning for Automatic Heuristic Design
AHD Agent trains a 4B-parameter LLM via agentic RL to actively use tools for automatic heuristic design, matching or exceeding larger baselines across eight domains with fewer evaluations.
-
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.
-
Weblica: Scalable and Reproducible Training Environments for Visual Web Agents
Weblica scales RL training for visual web agents by building thousands of reproducible environments through HTTP caching for stable replays and LLM synthesis from real sites, yielding an 8B model that beats similar open baselines on navigation benchmarks.
-
Back to the Beginning of Heuristic Design: Bridging Code and Knowledge with LLMs
A knowledge-first approach to LLM-driven automatic heuristic design in combinatorial optimization yields better discovery efficiency, transfer, and generalization than code-centric baselines by formalizing a distortion-compression trade-off.
-
Out-of-the-Box Global Optimization for Packing Problems: New Models and Improved Solutions
New nonlinear formulations for geometric packing problems, solved with FICO Xpress and SCIP, produce improved and first-known solutions for several variants.
-
Agentic-imodels: Evolving agentic interpretability tools via autoresearch
Agentic-imodels evolves scikit-learn regressors via an autoresearch loop to jointly boost predictive performance and LLM-simulatability, improving downstream agentic data science tasks by up to 73% on the BLADE benchmark.
-
A revision of Litvak's conjecture on Gaussian minima and a volumetric zone conjecture
Litvak's conjecture on minimizing moments of Gaussian minima is disproved by a cosine-based correlation matrix for small n and p, with a new conjecture proposed that this matrix is the general minimizer, supported conditionally via a volumetric extension of Fejes Tóth's zone conjecture.
-
BIM Information Extraction Through LLM-based Adaptive Exploration
LLM adaptive exploration via runtime code execution outperforms static query generation for information extraction from heterogeneous BIM models on the new ifc-bench v2 benchmark.
-
Agentic Harness Engineering: Observability-Driven Automatic Evolution of Coding-Agent Harnesses
AHE automates coding-agent harness evolution via component, experience, and decision observability, raising Terminal-Bench 2 pass@1 from 69.7% to 77.0% with transfer gains across models and benchmarks.
-
AI scientists produce results without reasoning scientifically
LLM agents execute scientific tasks but fail to follow core scientific reasoning norms such as evidence consideration and belief revision based on refutations.
-
Neuro-Symbolic ODE Discovery with Latent Grammar Flow
Latent Grammar Flow discovers ODEs by placing grammar-based equation representations in a discrete latent space, using a behavioral loss to cluster similar equations, and sampling via a discrete flow model guided by data fit and constraints.
-
Optimal and Near-Optimal Constructions for Bootstrap Percolation in Hypercubes
m(Q_d;4) equals d(d² + 3d + 14)/24 + 1 for infinitely many d, with an O(d)-additive upper bound for all d.
-
BEAM: Bi-level Memory-adaptive Algorithmic Evolution for LLM-Powered Heuristic Design
BEAM reformulates LLM-based heuristic design as bi-level optimization using GA for structures, MCTS for placeholders, and adaptive memory to outperform prior single-layer methods on CVRP and MIS tasks.
-
The AI Telco Engineer: Toward Autonomous Discovery of Wireless Communications Algorithms
An LLM-powered agentic framework autonomously designs competitive and sometimes superior explainable algorithms for wireless PHY and MAC layer tasks.
-
A Minimal Model of Representation Collapse: Frustration, Stop-Gradient, and Dynamics
A minimal embedding model shows representation collapse arises from frustrated samples through slow dynamics and is prevented by stop-gradient.
-
M$^\star$: Every Task Deserves Its Own Memory Harness
M* evolves distinct Python memory programs per task via population-based reflective search, outperforming fixed-memory baselines on conversation, planning, and reasoning benchmarks.
-
$k$-server-bench: Automating Potential Discovery for the $k$-Server Conjecture
k-server-bench formulates potential-function discovery for the k-server conjecture as a code-based inequality-satisfaction task; current agents fully solve the resolved k=3 case and reduce violations on the open k=4 case.
-
Autopoiesis: A Self-Evolving System Paradigm for LLM Serving Under Runtime Dynamics
Autopoiesis uses LLM-driven program synthesis to evolve serving policies online during deployment, delivering up to 53% and average 34% gains over prior LLM serving systems under runtime dynamics.
-
AutoSOTA: An End-to-End Automated Research System for State-of-the-Art AI Model Discovery
AutoSOTA uses eight specialized agents to replicate and optimize models from recent AI papers, producing 105 new SOTA results in about five hours per paper on average.
-
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.
-
Meta-Harness: End-to-End Optimization of Model Harnesses
Meta-Harness discovers improved harness code for LLMs via agentic search over prior execution traces, yielding 7.7-point gains on text classification with 4x fewer tokens and 4.7-point gains on math reasoning across held-out models.
-
Epistemic Uncertainty for Test-Time Discovery
UG-TTT adds epistemic uncertainty measured by adapter disagreement as an exploration bonus in RL for LLMs, raising maximum reward and diversity on scientific discovery benchmarks.
-
What should post-training optimize? A test-time scaling law perspective
Tail-extrapolated estimators approximate best-of-N policy gradients from limited training rollouts by leveraging upper-tail reward statistics under structural assumptions.
-
LLM4Branch: Large Language Model for Discovering Efficient Branching Policies of Integer Programs
LLM4Branch discovers branching policies for MILP solvers as LLM-generated executable programs whose parameters are tuned via zeroth-order optimization on solver performance.
-
SCALAR: A Neurosymbolic Framework for Automated Conjecture and Reasoning in Quantum Circuit Analysis
SCALAR generates conjectures linking optimal QAOA parameters to graph invariants, recovers known periodicity and parameter-transfer properties, and identifies correlations with optimization landscapes across thousands of graphs up to 77 qubits.
-
Metal-Sci: A Scientific Compute Benchmark for Evolutionary LLM Kernel Search on Apple Silicon
Metal-Sci is a benchmark and harness for LLM evolutionary optimization of Apple Silicon Metal kernels that uses held-out sizes to detect silent regressions missed by in-distribution scores.
-
MLS-Bench: A Holistic and Rigorous Assessment of AI Systems on Building Better AI
MLS-Bench shows that current AI agents fall short of reliably inventing generalizable ML methods, with engineering tuning easier than genuine invention.
-
FlashEvolve: Accelerating Agent Self-Evolution with Asynchronous Stage Orchestration
FlashEvolve accelerates LLM agent self-evolution via asynchronous stage orchestration and inspectable language-space staleness handling, reporting 3.5-4.9x proposal throughput gains over synchronous baselines on GEPA workloads.
-
Open-Ended Task Discovery via Bayesian Optimization
Generate-Select-Refine is an open-ended Bayesian optimization method that generates tasks and concentrates evaluations on the best one with only logarithmic regret overhead relative to standard single-task optimization.
-
Intentmaking and Sensemaking: Human Interaction with AI-Guided Mathematical Discovery
Expert mathematicians using an AI coding agent for discovery engage in repeated cycles of intentmaking to define goals and sensemaking to interpret outputs.
-
CodeEvolve: LLM-Driven Evolutionary Optimization with Runtime-Enriched Target Selection for Multi-Language Code Enhancement
CodeEvolve uses runtime-guided target selection and MCTS-augmented LLM evolution to optimize real Java and Apex code, reporting 15.22x average speedup on seven hotspots while preserving correctness.
-
CASCADE: Case-Based Continual Adaptation for Large Language Models During Deployment
CASCADE enables LLMs to continually adapt at deployment via case-based episodic memory and contextual bandits, improving macro-averaged success by 20.9% over zero-shot on 16 tasks spanning medicine, law, code, and robotics.
-
Programmatic Context Augmentation for LLM-based Symbolic Regression
Programmatic context augmentation lets LLM-based symbolic regression perform code-driven data analysis during search, yielding superior efficiency and accuracy over baselines on LLM-SRBench.
-
FitText: Evolving Agent Tool Ecologies via Memetic Retrieval
FitText embeds memetic evolutionary retrieval inside the agent's reasoning loop to iteratively refine pseudo-tool descriptions, raising retrieval rank from 8.81 to 2.78 on ToolRet and pass rate to 0.73 on StableToolBench.
-
Rethinking Network Topologies for Cost-Effective Mixture-of-Experts LLM Serving
Switchless topologies such as 3D full-mesh are 20.6-56.2% more cost-effective than scale-up networks for MoE LLM serving, with current link bandwidths over-provisioned by up to 27%.
-
CTM-AI: A Blueprint for General AI Inspired by a Model of Consciousness
CTM-AI combines a formal consciousness model with foundation models to report state-of-the-art results on sarcasm detection, humor, and agentic tool-use benchmarks.
-
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.
-
GR-Evolve: Design-Adaptive Global Routing via LLM-Driven Algorithm Evolution
GR-Evolve applies LLM-driven code evolution to global routing, reporting up to 8.72% post-detailed-routing wirelength reduction on seven benchmarks across three technology nodes.