# Сверяване

Автоматизирано съгласуване на счетоводни книги с агенти на езикови модели

11 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

12 юли 2026 г. 5 минути четене

## [FinRAGBench-V: Multimodal RAG with Visual Citations in the Financial Domain](https://beancount.io/bg/bean-labs/research-logs/2026/07/12/finragbench-v-multimodal-rag-visual-citation-financial-domain)

FinRAGBench-V finds top models reach only 20–61% block-level citation recall on financial pages. Multimodal retrieval beats text-only by nearly 50 points.

Mike Thrift

ai llm machine-learning

11 юли 2026 г. 6 минути четене

## [Can LLM Agents Be CFOs? EnterpriseArena's 132-Month Simulation Reveals a Wide Gap](https://beancount.io/bg/bean-labs/research-logs/2026/07/11/can-llm-agents-be-cfos-enterprisearena-resource-allocation-benchmark)

Only Qwen3.5-9B survives 80% of EnterpriseArena's 132-month CFO runs, while GPT-5.4 and DeepSeek-V3.1 hit 0%. Skipped ledger reconciliation causes the failures.

Mike Thrift

ai llm automation

7 юли 2026 г. 5 минути четене

## [FinMCP-Bench: Benchmarking LLM Agents for Real-World Financial Tool Use under MCP](https://beancount.io/bg/bean-labs/research-logs/2026/07/07/finmcp-bench-llm-agents-financial-tool-use-model-context-protocol)

FinMCP-Bench scores the best of six LLMs at just 3.08% exact match on 613 real MCP financial tasks, a 20× collapse from single-tool to multi-turn use.

Mike Thrift

ai llm automation

2 юли 2026 г. 6 минути четене

## [Found in the Middle: Calibrating Positional Attention Bias Improves Long-Context RAG](https://beancount.io/bg/bean-labs/research-logs/2026/07/02/found-in-the-middle-calibrating-positional-attention-bias)

Subtracting positional bias from LLM attention weights recovers up to 15 points of RAG accuracy when evidence sits mid-context, aiding finance agent pipelines.

Mike Thrift

ai llm machine-learning

29 юни 2026 г. 6 минути четене

## [Fin-RATE: How LLMs Fail at Cross-Period and Cross-Entity Financial Analysis](https://beancount.io/bg/bean-labs/research-logs/2026/06/29/fin-rate-real-world-financial-analytics-tracking-evaluation-benchmark)

Fin-RATE shows LLM accuracy collapses 18.60% on longitudinal tracking, with the retrieval pipeline, not the model, as the binding bottleneck.

Mike Thrift

llm ai machine-learning

8 май 2026 г. 6 минути четене

## [Voyager: skill libraries unlock 3.3× more item discovery](https://beancount.io/bg/bean-labs/research-logs/2026/05/08/voyager-open-ended-embodied-agent-lifelong-learning)

Voyager's persistent code skill library discovers 3.3× more Minecraft items than prior SOTA without fine-tuning—the reuse pattern ledger agents need.

Mike Thrift

ai llm machine-learning

4 май 2026 г. 5 минути четене

## [AutoGen: Multi-Agent Conversation Frameworks for Finance AI](https://beancount.io/bg/bean-labs/research-logs/2026/05/04/autogen-multi-agent-conversation-framework)

AutoGen's two-agent conversation lifts MATH accuracy from 55% to 69%, and its SafeGuard agent adds up to 35 F1 points on unsafe-code detection.

Mike Thrift

ai llm automation

29 април 2026 г. 5 минути четене

## [CodeAct: Why Executable Python Code Makes LLM Agents 20% More Accurate](https://beancount.io/bg/bean-labs/research-logs/2026/04/29/codeact-executable-code-actions-llm-agents)

CodeAct replaces JSON tool calls with executable Python, lifting GPT-4 agent success by \~20 points and cutting turns 30%.

Mike Thrift

ai llm automation

26 април 2026 г. 5 минути четене

## [CRITIC: Why LLM Self-Correction Requires External Tool Feedback](https://beancount.io/bg/bean-labs/research-logs/2026/04/26/critic-llm-self-correct-tool-interactive-critiquing)

CRITIC lifts open-domain QA by 7.7 F1 and cuts toxicity 79.2% by verifying LLM revisions against external tools, not itself.

Mike Thrift

ai llm machine-learning

17 април 2026 г. 6 минути четене

## [ReAct: Синергия между разсъждението и действието в езиковите модели](https://beancount.io/bg/bean-labs/research-logs/2026/04/17/react-synergizing-reasoning-and-acting-in-language-models)

ReAct редува разсъждение с действия с инструменти, побеждавайки чистия CoT с 34 точки при проверка на факти. Провалите му оформят агентите, пишещи в регистри на Beancount.

Mike Thrift

ai llm machine-learning

[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
