# Счетоводство с обикновен текст

Изследвания, основани на счетоводни формати и работни процеси в обикновен текст

34 публикации [Вижте всички етикети](https://beancount.io/bg/bean-labs/research-logs/tags) [Обратно към всички публикации](https://beancount.io/bg/bean-labs/research-logs)

10 септември 2026 г. 6 минути четене

## [Може ли вашият агент да изравни тези сметки?](https://beancount.io/bg/bean-labs/research-logs/2026/09/10/can-your-agent-balance-these-books)

Едно изпълнение на агент осчетоводи книга с три реда до 2958.50 USD разплащателна сметка и 1958.50 USD септемврийска печалба, след като сам диагностицира провал.

Mike Thrift

agents llm beancount

1 юли 2026 г. 6 минути четене

## [Uncertainty-Aware Deferral for LLM Agents: When to Escalate from Small to Large Models](https://beancount.io/bg/bean-labs/research-logs/2026/07/01/redact-uncertainty-aware-deferral-llm-agents)

ReDAct defers from a small model to a large one only when perplexity signals uncertainty, cutting cost 64% at matching accuracy for agent workflows.

Mike Thrift

ai llm automation

30 юни 2026 г. 6 минути четене

## [OpenHands: Open Platform for AI Software Agents and What It Means for Finance Automation](https://beancount.io/bg/bean-labs/research-logs/2026/06/30/openhands-open-platform-ai-software-developers-generalist-agents)

OpenHands' CodeAct agent scores 26% on SWE-Bench Lite, showing what AI agents reliably do today. Finance automation should start tightly scoped, not autonomous.

Mike Thrift

ai open-source automation

23 юни 2026 г. 6 минути четене

## [LLMs Score 2.3% on Beancount DSL Generation: The LLMFinLiteracy Benchmark](https://beancount.io/bg/bean-labs/research-logs/2026/06/23/llm-beancount-dsl-financial-literacy-benchmark)

LLMFinLiteracy finds five \~7B models write correct Beancount transactions just 2.3% of the time, failing on accounting reasoning rather than syntax.

Mike Thrift

llm beancount plain-text-accounting

22 юни 2026 г. 6 минути четене

## [TableMaster: Adaptive Reasoning for Table Understanding with LLMs](https://beancount.io/bg/bean-labs/research-logs/2026/06/22/tablemaster-adaptive-reasoning-table-understanding)

TableMaster hits 78.13% on WikiTQ with GPT-4o-mini, 13 points over Chain-of-Table, via table-of-focus plus adaptive reasoning for agents over Beancount ledgers.

Mike Thrift

ai llm machine-learning

18 юни 2026 г. 6 минути четене

## [τ²-bench: Measuring the Cost of Dual-Control in Conversational AI Agents](https://beancount.io/bg/bean-labs/research-logs/2026/06/18/tau-squared-bench-dual-control-conversational-agents)

τ²-bench finds active users with their own tools cut conversational agent success by 18–25 points. Beancount agents sharing write access lose the same way.

Mike Thrift

ai llm automation

16 юни 2026 г. 6 минути четене

## [GAIA Benchmark: Measuring What Frontier AI Agents Can Actually Do](https://beancount.io/bg/bean-labs/research-logs/2026/06/16/gaia-benchmark-general-ai-assistants)

GAIA's 466 tasks show frontier AI agents at 74.55% versus 92% for humans, with Level 3 coordination still the hardest gap to close.

Mike Thrift

ai llm machine-learning

13 юни 2026 г. 6 минути четене

## [WorkArena: How LLM Web Agents Perform on Real Enterprise Knowledge Work](https://beancount.io/bg/bean-labs/research-logs/2026/06/13/workarena-web-agents-enterprise-knowledge-work-tasks)

WorkArena's 33 ServiceNow tasks show GPT-4o at 42.7% overall but 0% on list filters, a wall structured UI interaction must clear.

Mike Thrift

ai llm automation

12 юни 2026 г. 6 минути четене

## [τ-bench: Measuring AI Agent Reliability in Real-World Tool-Use Domains](https://beancount.io/bg/bean-labs/research-logs/2026/06/12/tau-bench-tool-agent-user-interaction-real-world-domains)

τ-bench finds top LLMs fall from pass\@1 0.692 to pass\@4 0.462 on retail tool-use tasks. Write-back agents on a ledger face the same consistency cliff.

Mike Thrift

ai llm machine-learning

11 юни 2026 г. 6 минути четене

## [Chain-of-Table: Evolving Tables in the LLM Reasoning Chain](https://beancount.io/bg/bean-labs/research-logs/2026/06/11/chain-of-table-evolving-tables-reasoning-chain)

Chain-of-Table hits 67.31% on WikiTQ versus 61.48% for text-only chain-of-thought, and leads by 10.25 points on tables over 4,000 tokens.

Mike Thrift

ai llm machine-learning

10 юни 2026 г. 5 минути четене

## [TableLlama: Can a 7B Open Model Match GPT-4 on Table Understanding?](https://beancount.io/bg/bean-labs/research-logs/2026/06/10/tablellama-open-generalist-models-tables)

TableLlama beats GPT-4 on column type annotation (F1 94 vs 32) but trails by 33 points on WikiTQ compositional reasoning.

Mike Thrift

llm ai machine-learning

9 юни 2026 г. 5 минути четене

## [TAPAS: Weakly Supervised Table QA Without SQL, and What It Means for Beancount](https://beancount.io/bg/bean-labs/research-logs/2026/06/09/tapas-weakly-supervised-table-parsing-pretraining)

TAPAS answers table questions by selecting cells, never generating SQL. It fits small Beancount ledger queries but breaks down at scale.

Mike Thrift

ai machine-learning llm

[Започни безплатно→](https://beancount.io/sign-up/?next=%2Fledger%2F) [Документация](https://beancount.io/bg/docs)

## Where to look next

- [Sitemap](https://beancount.io/sitemap.xml) - all pages with hreflang alternates
- [LLMs index](https://beancount.io/llms.txt) - curated links for AI agents (also /llms-full.txt for full docs)
- [Documentation](https://beancount.io/docs) - every doc is also available as text/markdown via .md suffix or Accept negotiation
- [REST API reference](https://beancount.io/docs/api-reference) - interactive endpoints, schemas, authentication, and code samples
- [OpenAPI spec](https://beancount.io/openapi.json) - OpenAPI 3.1 spec for this website's public content endpoints; the product contract for ledger operations is at https://beancount.io/api-gateway/v1/openapi.json
- [GraphQL](https://beancount.io/graphql) - public, read-only content API over the blog and research logs (no credentials, no ledger data): 12 query fields over 23 types
- [MCP server](https://beancount.io/mcp) - Model Context Protocol at https://beancount.io/api-gateway/mcp
