Absorbing Complexity: An Interaction-Native Knowledge Harness for Financial LLM Agents
Summary
Proposes InKH, an interaction-native knowledge harness architecture for financial LLM agents that absorbs user complexity through structured knowledge management and temporal memory, achieving significant improvements in latency, token cost, and stale-knowledge reduction.
View Cached Full Text
Cached at: 06/05/26, 02:08 PM
Paper page - Absorbing Complexity: An Interaction-Native Knowledge Harness for Financial LLM Agents
Source: https://huggingface.co/papers/2606.01886 Published on Jun 1
·
Submitted byhttps://huggingface.co/pidturkina
Sofiaon Jun 5
Abstract
Financial AI agents struggle with user complexity, but a new architecture called InKH addresses this by embedding complexity into the system through structured knowledge management and temporal memory mechanisms.
Financial AI agents often fail for a simple reason: they make users carry the complexity. A user must repeatedly restate goals, risk preferences, portfolio context, past judgments, and shifting market assumptions, while the agent answers, retrieves, acts, and forgets. In finance, this is not just inconvenient. In tasks such as market analysis, copy-trading review, and trade preparation, forgotten context and stale memory can create latency, repeated errors, weak auditability, and unsafe decisions. We propose the interaction-nativeknowledge harness(InKH), an architecture forfinancial LLM agentsthat absorbs complexity into the system. InKH converts user, market, portfolio, and tool events into structured operational knowledge. It usespassive knowledge injectionto assemble a bounded working context buffer before the main model step,temporal graph memoryfor low-latency retrieval, awiki audit surfacefor human-readable governance, andbackground extractionwithmaturity,decay, andwrite-time invalidation. We evaluate InKH on a reproducible controlled synthetic benchmark with 24 random seeds, 4 rounds, 80 episodes per round, and 6 baselines, producing 46,080 baseline-conditioned evaluations. InKH achieves mean task quality of 0.815 at 900 ms latency. Compared with agent-driven wiki-walk memory, it reduces latency by 82.95 percent, token cost by 82.29 percent, and stale-knowledge usage by 96.58 percent, while improving quality by 0.108 and traceability by 0.461. Compared with a temporal-graph system without invalidation, it improves quality by 0.050 and reduces stale-memory usage by 96.58 percent with comparable serving cost. The results support a design thesis for financial AI: adoption happens when complexity is absorbed by the system rather than transferred to the user. The benchmark validates architecture-level behavior, not live trading performance.
View arXiv pageView PDFAdd to collection
Get this paper in your agent:
hf papers read 2606\.01886
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2606.01886 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2606.01886 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2606.01886 in a Space README.md to link it from this page.
Collections including this paper1
Similar Articles
Pushing the Limits of LLM Tool Calling via Experiential Knowledge Integration and Activation
This paper presents a systematic study on integrating experiential knowledge into LLM tool calling, proposing the KATE framework that combines knowledge-augmented data, width-expanded inference, and knowledge-aware training, achieving consistent improvements on BFCL-V3 and AppWorld benchmarks.
Ignorance or Incompetence? Constructing Knowledge-Gated, Verifiable Tasks for LLM Agents
The paper introduces a knowledge-gated task-construction protocol to explicitly test LLM agents' dependence on hidden knowledge, validated through calibration tasks showing performance drops without access to private conventions.
KC-Bench: A Dynamic Interactive Benchmark for Evaluating Knowledge Conflicts in LLM Agents
KC-Bench is a dynamic interactive benchmark for evaluating how LLM agents detect and resolve knowledge conflicts between user instructions, parametric knowledge, and environmental observations, with evaluations showing no model handles all conflict types reliably.
Beyond LLMs: Why Scalable Enterprise AI Adoption Depends on Agent Logic
IBM Research explores how agent logic—software primitives like knowledge graphs and program analysis—can guide LLM-based agents to efficiently handle complex enterprise workflows, reducing hallucinations and costs while improving outcomes.
LLM Wiki v2 (16 minute read)
This post presents a pattern for building personal knowledge bases using LLMs, offering a structured approach for leveraging large language models in knowledge management.