Latest research

Jun 25, 2026

5 mins

A definition of good explanations and the challenges explaining LLM outputs

Current neural networks struggle to provide satisfying explanations not due to a lack of information, but because their internal representations cannot be broken down into the distinct, independent facts required for human understanding, hence the "explainability paradox". The authors propose a definition of a good explanation based on counterfactuals that the recipient finds surprising (low prior probability), and suggest that achieving true explainability in AI will likely require integrating symbolic methods with neural systems to allow for decomposable information.

Louis Mahon

Jun 25, 2026

5 mins

A definition of good explanations and the challenges explaining LLM outputs

Current neural networks struggle to provide satisfying explanations not due to a lack of information, but because their internal representations cannot be broken down into the distinct, independent facts required for human understanding, hence the "explainability paradox". The authors propose a definition of a good explanation based on counterfactuals that the recipient finds surprising (low prior probability), and suggest that achieving true explainability in AI will likely require integrating symbolic methods with neural systems to allow for decomposable information.

Louis Mahon

Jun 25, 2026

5 mins

A definition of good explanations and the challenges explaining LLM outputs

Current neural networks struggle to provide satisfying explanations not due to a lack of information, but because their internal representations cannot be broken down into the distinct, independent facts required for human understanding, hence the "explainability paradox". The authors propose a definition of a good explanation based on counterfactuals that the recipient finds surprising (low prior probability), and suggest that achieving true explainability in AI will likely require integrating symbolic methods with neural systems to allow for decomposable information.

Louis Mahon

Jun 25, 2026

5 mins

A definition of good explanations and the challenges explaining LLM outputs

Current neural networks struggle to provide satisfying explanations not due to a lack of information, but because their internal representations cannot be broken down into the distinct, independent facts required for human understanding, hence the "explainability paradox". The authors propose a definition of a good explanation based on counterfactuals that the recipient finds surprising (low prior probability), and suggest that achieving true explainability in AI will likely require integrating symbolic methods with neural systems to allow for decomposable information.

Louis Mahon

Jun 25, 2026

5 mins

A definition of good explanations and the challenges explaining LLM outputs

Current neural networks struggle to provide satisfying explanations not due to a lack of information, but because their internal representations cannot be broken down into the distinct, independent facts required for human understanding, hence the "explainability paradox". The authors propose a definition of a good explanation based on counterfactuals that the recipient finds surprising (low prior probability), and suggest that achieving true explainability in AI will likely require integrating symbolic methods with neural systems to allow for decomposable information.

Louis Mahon

May 11, 2026

2 min

On attention sinks: What we found in LLMs

An interesting observed behaviour of LLMs is that they sometimes place nearly all their attention onto a single meaningless token. Why this happens is not fully understood. New research uncovers a smaller, subtler variant that appears mid-generation, and it's another sign that LLMs decide what matters in ways that look nothing like human reasoning.

Louis Mahon

May 11, 2026

2 min

On attention sinks: What we found in LLMs

An interesting observed behaviour of LLMs is that they sometimes place nearly all their attention onto a single meaningless token. Why this happens is not fully understood. New research uncovers a smaller, subtler variant that appears mid-generation, and it's another sign that LLMs decide what matters in ways that look nothing like human reasoning.

Louis Mahon

May 11, 2026

2 min

On attention sinks: What we found in LLMs

An interesting observed behaviour of LLMs is that they sometimes place nearly all their attention onto a single meaningless token. Why this happens is not fully understood. New research uncovers a smaller, subtler variant that appears mid-generation, and it's another sign that LLMs decide what matters in ways that look nothing like human reasoning.

Louis Mahon

May 11, 2026

2 min

On attention sinks: What we found in LLMs

An interesting observed behaviour of LLMs is that they sometimes place nearly all their attention onto a single meaningless token. Why this happens is not fully understood. New research uncovers a smaller, subtler variant that appears mid-generation, and it's another sign that LLMs decide what matters in ways that look nothing like human reasoning.

Louis Mahon

May 11, 2026

2 min

On attention sinks: What we found in LLMs

An interesting observed behaviour of LLMs is that they sometimes place nearly all their attention onto a single meaningless token. Why this happens is not fully understood. New research uncovers a smaller, subtler variant that appears mid-generation, and it's another sign that LLMs decide what matters in ways that look nothing like human reasoning.

Louis Mahon

Apr 24, 2026

2 mins

Neurosymbolic AI explained (with legislation example)

Why do LLMs fail on complex logic? Discover the structural limits of deep learning and how the neurosymbolic principle provides explainable, consistent, and trustworthy AI for regulated industries.

Callum Hackett

Apr 24, 2026

2 mins

Neurosymbolic AI explained (with legislation example)

Why do LLMs fail on complex logic? Discover the structural limits of deep learning and how the neurosymbolic principle provides explainable, consistent, and trustworthy AI for regulated industries.

Callum Hackett

Apr 24, 2026

2 mins

Neurosymbolic AI explained (with legislation example)

Why do LLMs fail on complex logic? Discover the structural limits of deep learning and how the neurosymbolic principle provides explainable, consistent, and trustworthy AI for regulated industries.

Callum Hackett

Apr 24, 2026

2 mins

Neurosymbolic AI explained (with legislation example)

Why do LLMs fail on complex logic? Discover the structural limits of deep learning and how the neurosymbolic principle provides explainable, consistent, and trustworthy AI for regulated industries.

Callum Hackett

Apr 24, 2026

2 mins

Neurosymbolic AI explained (with legislation example)

Why do LLMs fail on complex logic? Discover the structural limits of deep learning and how the neurosymbolic principle provides explainable, consistent, and trustworthy AI for regulated industries.

Callum Hackett

Apr 24, 2026

2 mins