{"id":23609,"date":"2025-07-09T12:54:48","date_gmt":"2025-07-09T12:54:48","guid":{"rendered":"https:\/\/staging.embraceable.ai\/ankuendigung\/e2-research1-entering-the-next-phase-of-cognitive-reasoning\/"},"modified":"2025-07-18T08:52:34","modified_gmt":"2025-07-18T08:52:34","slug":"e2-research1-entering-the-next-phase-of-cognitive-reasoning","status":"publish","type":"ankuendigung","link":"https:\/\/new-staging.easy-dashboard.de\/en\/ankuendigung\/e2-research1-entering-the-next-phase-of-cognitive-reasoning\/","title":{"rendered":"e2-research#1: Entering the Next Phase of Cognitive Reasoning"},"content":{"rendered":"\t\t
\n\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

The First Breakthrough: e1<\/h4>

In the first half of 2025, with e1<\/a>, we presented the first functional breakthrough of our new foundational architecture for Cognitive Reasoning<\/strong>. For the first time, we extended a language model with a cognitive control component to form what we call a “Cognitive System.<\/strong>” Even e1 impressively demonstrated that deductive reasoning and reconstructible argumentation paths can be mastered within the scope of machine text processing and generation\u2014a pioneering step towards controllable, explainable, and simultaneously high-performance AI. <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

What is Cognitive Reasoning?<\/h4>

Cognitive Reasoning is a fundamentally new architectural<\/strong> approach for high-performance and simultaneously controllable AI systems. This architecture combines the expressive power and semantic diversity of large language models (LLMs) with a formal control unit: the Cognitive Control Unit<\/strong> (CCU). The CCU structures and directs thought processes based on so-called Cognitive Schemata<\/strong> and auditable logic artifacts (axioms). Thus, the CCU determines how <\/em>thoughts <\/span>are formed<\/strong>, not just what is said<\/em><\/strong>. This creates traceable, explicit thought paths that give the overall process structure and control without curtailing the actual intelligence. <\/p>

For a detailed description of the architecture, we refer to our technical article:
“Cognitive Reasoning \u2013 an innovative foundational architecture for controllable, structured AI systems.<\/u>“<\/em><\/a><\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

The goal of this present e2-research#1<\/span> publication is to provide initial answers to the question: how far can this architecture take us?<\/h4>

Following the promising results of e1, the central question quickly arose: can the (benchmark) performance be further increased\u2014and can this be done solely <\/em>through the optimization of the cognitive architecture (i.e., without resource-intensive and time-consuming training)? We have dedicated ourselves to this question within the scope of this publication. <\/p>

The following analyses (e2-research#1) are based on initial studies that we conducted over a two-week period in the second half of June 2025 in search of answers to the aforementioned question. The underlying language models remained unchanged in all analyses; only the architecture of the CCU (more specifically: the Cognitive Schemata used) was modified\u2014with the aim of making the thought structure even more efficient, consistent, and adaptive. <\/p>

For reasons of speed and validation, the analyses presented below were conducted in parallel <\/span>with two different base models: an open-weight Llama 3.3 70B and the closed-weight GPT-4.1-mini. For the upcoming publication (e2-research#2), both benchmarks will also be performed with additional base models. <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t\t\t
\n\t\t\t\n\t\t\t\t\t\t<\/span>\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Analysis Dimension 1<\/span>: Logical Puzzles with Semantic Components (HardReasoning-DE)<\/h4>

The HardReasoning-DE benchmark tests AI systems with challenging logical puzzles that go far beyond simple conclusions. The tasks combine structured logic with complex semantic contexts<\/strong>\u2014for example, analyzing sports team strategies over several games or optimizing project resources under sustainability constraints. <\/p>

These challenges require not only logical thinking but also the ability to understand semantic constraints, set priorities<\/strong>, and develop consistent solution paths in combination with the logical requirements. This combination of logic and semantics is less common in the training data of language and reasoning models than pure logic puzzles; therefore, the ability for dynamic, complex reasoning (as opposed to the mere reproduction of pre-trained language patterns) plays a particularly significant role here. <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Analysis Dimension 1<\/span>: Logical Puzzles with Semantic Components (HardReasoning-DE)<\/h4>

The HardReasoning-DE benchmark tests AI systems with challenging logical puzzles that go far beyond simple conclusions. The tasks combine structured logic with complex semantic contexts<\/strong>\u2014for example, analyzing sports team strategies over several games or optimizing project resources under sustainability constraints. <\/p>

These challenges require not only logical thinking but also the ability to understand semantic constraints, set priorities<\/strong>, and develop consistent solution paths in combination with the logical requirements. This combination of logic and semantics is less common in the training data of language and reasoning models than pure logic puzzles; therefore, the ability for dynamic, complex reasoning (as opposed to the mere reproduction of pre-trained language patterns) plays a particularly significant role here. <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

HardReasoning-DE Results:<\/u><\/h4>

e2 shows a significant performance leap compared to e1 in correctly solving the puzzles. Both systems use a medium-sized Llama 3.3 70B as a hypothesis generator and validator for this benchmark. <\/p>

Even in this setup, e2 surpasses leading reasoning models like o3-mini and is only narrowly beaten by o4-mini.<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t\t\t
\n\t\t\t\n\t\t\t\t\t\t<\/span>\n\t\t<\/div>\n\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

Analysis Dimension 2<\/span>: Purely Logical Puzzles with High Logical Depth (ZebraLogicBench)<\/span><\/h4>

To evaluate primarily logical performance, we use the established benchmark ZebraLogicBench <\/strong><\/a>from AllenAI (Ai2) & Stanford University, which investigates the scaling limits of LLMs in logical reasoning. The benchmark measures the ability of models to maintain consistent thinking operations even across multiple deductive steps\u2014an ideal reference for our goal. <\/p>

These tasks follow the same principle as the famous Einstein Riddle (also known as the “Zebra Puzzle”<\/a>), where people with different characteristics are assigned to different houses. For illustrative purposes, this Einstein Riddle is shown below: <\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

For the initial analyses in the context of the e2-research#1 investigation, we limited ourselves to the XL variants of the Zebra benchmark for time reasons. These XL variants are particularly challenging as they require a larger number of variables and constraints, as well as significantly longer chains of logical conclusions. The processing time for humans is typically between 30 and 45 minutes. The XL puzzles range in complexity from 4 houses with 6 attributes to the most difficult puzzles with a 6×6 matrix. <\/p>

For our evaluation, we translated the puzzles of the ZebraLogicBench into German <\/strong>to ensure correct linguistic reference.<\/p>

The initial analysis results are presented in the following table:<\/p>\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t

\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\n\t\t\t\t\t\t\t\"\"\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t
\n\t\t\t\t
\n\t\t\t\t\t\t\t\t\t

ZebraLogicBench Results: The Influence of Cognitive Architecture<\/h4>

The initial analysis results impressively show a) the influence <\/strong>and b) the progress <\/strong>of the cognitive architecture: While e1@gpt-4.1-mini achieves 4th place in the benchmark, e2-research@gpt-4.1-mini, with its stronger ‘Cognitive Schema,’ is currently positioned at 2nd place in the rankings (as of July 10, 2025)<\/em>. This significant improvement was achieved without any changes to the underlying language model\u2014solely through the systematic further development of the CCU architecture (specifically, the ‘Cognitive Schema’). <\/p>

In direct comparison, the performance gains are clearly visible:<\/p>