Experience Graphs: The Data Foundation for Self-Improving Agents

• URL: http://arxiv.org/abs/2606.29823v1
• Date: Mon, 29 Jun 2026 06:02:20 GMT
• Title: Experience Graphs: The Data Foundation for Self-Improving Agents
• Abstract summary: We argue that long-horizon agentic tasks -- code generation, scientific discovery, design -- are such a workload.<n>This search produces a structured object we call an experience graph.<n>We propose Trellis: a data foundation that treats the experience graph as first-class, governed, queryable database.
• Score: 24.60392207371798
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: The database community has repeatedly advanced the state of the art by recognizing that new workloads demand new system architectures. We argue that long-horizon agentic tasks -- code generation, scientific discovery, hardware design -- are such a workload. These agents explore: they generate artifacts, execute tools, observe failures, branch, and repair over hundreds of steps. This search produces a structured object we call an experience graph: executable artifacts, tool outputs, rewards, sibling comparisons, and causal lineage. Yet existing agent frameworks treat this experience as disposable state -- JSON checkpoints and session logs that cannot be recovered after a crash, queried across users, or materialized into training data. We propose Trellis: a data foundation that treats the experience graph as first-class, governed, queryable database state. The core insight is that search over experience graphs is a database access pattern. Frontier selection is a query, cross-session reuse is vector-seeded graph retrieval, training-data extraction is a materialized view, and reconstructing what an agent knew at any past step is a time-travel query. When the database owns the experience graph, agents become stateless compute, and crash recovery, horizontal scaling, and a closed-loop training flywheel emerge as architectural byproducts. We ground the design in KernelEvolve, a production accelerator-kernel optimizer at Meta, where cross-session reuse reaches a target speedup roughly 10x faster at 52% lower token cost. More broadly, Trellis turns inference-time search from disposable computation into a durable institutional asset: logs made databases reliable; experience graphs may make agents cumulative.

Related papers

• From Final Artifacts to Trajectories: Retrospective Process Supervision for Evidence-Grounded Long-Form Generation [49.496216822640974]
  RetroGen is a self-improving framework for retrospective process supervision.<n>It reconstructs latent trajectories from expert artifacts, verifies them against the artifact and supporting evidence, and trains models on their own successful reconstruction data.<n>Experiments show that RetroGen improves grounding, faithful synthesis, and long-form evidence-seeking agent tasks.
  arXiv  Detail & Related papers  (2026-08-31T08:47:39Z)
• AgentTrails: Towards Trust and Reuse for Agentic Tasks [12.615897924172812]
  LLM-powered agents tackle complex tasks by invoking tools, querying databases, executing code, and manipulating intermediate artifacts.<n>These agents follow trajectories that are typically stored as chronological logs, obscuring the underlying dataflow.<n>We present AgentTrails, a prototype system for agent provenance and sensemaking.
  arXiv  Detail & Related papers  (2026-07-21T07:53:58Z)
• SEGRA: Structured Experience-Guided Graph Reasoning Agent for Gremlin Based Question Answering [1.4924423820656074]
  We introduce SEGRA, an experience-guided agent for text-to-Gremlin question answering.<n>On an enterprise IT support benchmark, SEGRA achieves a $7.0times$ higher mean judge score than backbone-only chain-of-thought prompting.
  arXiv  Detail & Related papers  (2026-07-21T03:22:40Z)
• The Log is the Agent: Event-Sourced Reactive Graphs for Auditable, Forkable Agentic Systems [0.0]
  We describe ActiveGraph, a runtime that inverts the typical agent framework.<n>The append-only event log is the source of truth; the working graph is a deterministic projection of that log.<n>No component instructs another; coordination happens entirely through the shared graph.
  arXiv  Detail & Related papers  (2026-05-21T04:55:38Z)
• GraphWalk: Enabling Reasoning in Large Language Models through Tool-Based Graph Navigation [0.2999888908665658]
  We present GraphWalk, a problem-agnostic, training-free, tool-based framework that allows off-the-shelf LLMs to reason through sequential graph navigation.<n>We evaluate whether models equipped with GraphWalk can compose these operations into correct multi-step reasoning chains.
  arXiv  Detail & Related papers  (2026-04-02T04:43:01Z)
• GraphER: An Efficient Graph-Based Enrichment and Reranking Method for Retrieval-Augmented Generation [47.014599006114594]
  GraphER is a graph-based enrichment and reranking method that captures multiple forms of proximity beyond semantic similarity.<n>It does not require a knowledge graph, allowing GraphER to integrate seamlessly with standard vector stores.<n> Experiments on multiple retrieval benchmarks demonstrate the effectiveness of the proposed approach.
  arXiv  Detail & Related papers  (2026-03-26T01:30:22Z)
• GraphRunner: A Multi-Stage Framework for Efficient and Accurate Graph-Based Retrieval [3.792463570467098]
  GraphRunner is a novel graph-based retrieval framework that operates in three distinct stages: planning, verification, and execution.<n>It significantly reduces reasoning errors and detects hallucinations before execution.<n>Our evaluation using the GRBench dataset shows that GraphRunner consistently outperforms existing approaches.
  arXiv  Detail & Related papers  (2025-07-11T18:10:01Z)
• AgentSwift: Efficient LLM Agent Design via Value-guided Hierarchical Search [58.98450205734779]
  Large language model (LLM) agents have demonstrated strong capabilities across diverse domains.<n>Existing agent search methods suffer from three major limitations.<n>We introduce a comprehensive framework to address these challenges.
  arXiv  Detail & Related papers  (2025-06-06T12:07:23Z)
• GraphReader: Building Graph-based Agent to Enhance Long-Context Abilities of Large Language Models [58.08177466768262]
  Long-context capabilities are essential for large language models (LLMs) to tackle complex and long-input tasks. We introduce GraphReader, a graph-based agent system designed to handle long texts by structuring them into a graph and employing an agent to explore this graph autonomously. Experimental results on the LV-Eval dataset reveal that GraphReader, using a 4k context window, consistently outperforms GPT-4-128k across context lengths from 16k to 256k by a large margin.
  arXiv  Detail & Related papers  (2024-06-20T17:57:51Z)
• A Unified Active Learning Framework for Annotating Graph Data with Application to Software Source Code Performance Prediction [4.572330678291241]
  We develop a unified active learning framework specializing in software performance prediction. We investigate the impact of using different levels of information for active and passive learning. Our approach aims to improve the investment in AI models for different software performance predictions.
  arXiv  Detail & Related papers  (2023-04-06T14:00:48Z)
• Neural Graph Reasoning: Complex Logical Query Answering Meets Graph Databases [63.96793270418793]
  Complex logical query answering (CLQA) is a recently emerged task of graph machine learning. We introduce the concept of Neural Graph Database (NGDBs) NGDB consists of a Neural Graph Storage and a Neural Graph Engine.
  arXiv  Detail & Related papers  (2023-03-26T04:03:37Z)
• Retrieval-Augmented Reinforcement Learning [63.32076191982944]
  We train a network to map a dataset of past experiences to optimal behavior. The retrieval process is trained to retrieve information from the dataset that may be useful in the current context. We show that retrieval-augmented R2D2 learns significantly faster than the baseline R2D2 agent and achieves higher scores.
  arXiv  Detail & Related papers  (2022-02-17T02:44:05Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.