What Happens Inside a Language Model
The Setup
Posts by Tag
- mechanistic-interpretability 4
- alignment 2
- stage-attribution 2
- talkie 2
- dpo 2
- developmental-interpretability 2
- alignment-faking 1
- interpretability 1
- SAE 1
- multi-agent 1
- truth-contagion 1
- sae 1
- ai-safety 1
- llm-safety 1
- prompt-injection 1
- sycophancy 1
- sft 1
- control-validation 1
- vpd 1
- weight-diff 1
- authority-bias 1
- circuit-analysis 1
mechanistic-interpretability
SVD and VPD: Two Methods for Reading What Training Wrote Into a Model’s Weights
May 6, 2026 — bigsnarfdude
DPO Does Different Things to Different Axes — Stage Attribution on talkie-1930-13b
Numbers from n=50 across base/SFT/IT on talkie-1930-13b and base/SFT/IT on OLMo-3-7B. All claims are Tier-1 (detector-level) unless explicitly noted. Recruit...
Chaos Takes the Wheel: Salience-Weighted Attentional Hijacking via True Statements
April 2026 — bigsnarfdude
alignment
What Happens Inside a Language Model
The Setup
Why AI Has a Split Personality (And How to Trigger the Evil Twin)
April 2026 — bigsnarfdude
stage-attribution
SVD and VPD: Two Methods for Reading What Training Wrote Into a Model’s Weights
May 6, 2026 — bigsnarfdude
DPO Does Different Things to Different Axes — Stage Attribution on talkie-1930-13b
Numbers from n=50 across base/SFT/IT on talkie-1930-13b and base/SFT/IT on OLMo-3-7B. All claims are Tier-1 (detector-level) unless explicitly noted. Recruit...
talkie
SVD and VPD: Two Methods for Reading What Training Wrote Into a Model’s Weights
May 6, 2026 — bigsnarfdude
DPO Does Different Things to Different Axes — Stage Attribution on talkie-1930-13b
Numbers from n=50 across base/SFT/IT on talkie-1930-13b and base/SFT/IT on OLMo-3-7B. All claims are Tier-1 (detector-level) unless explicitly noted. Recruit...
dpo
What Happens Inside a Language Model
The Setup
DPO Does Different Things to Different Axes — Stage Attribution on talkie-1930-13b
Numbers from n=50 across base/SFT/IT on talkie-1930-13b and base/SFT/IT on OLMo-3-7B. All claims are Tier-1 (detector-level) unless explicitly noted. Recruit...
developmental-interpretability
SVD and VPD: Two Methods for Reading What Training Wrote Into a Model’s Weights
May 6, 2026 — bigsnarfdude
DPO Does Different Things to Different Axes — Stage Attribution on talkie-1930-13b
Numbers from n=50 across base/SFT/IT on talkie-1930-13b and base/SFT/IT on OLMo-3-7B. All claims are Tier-1 (detector-level) unless explicitly noted. Recruit...
alignment-faking
Detecting Alignment Faking in Language Models
Welcome to my new site. I’ll be writing about AI safety research, alignment faking detection, and whatever else seems interesting.
interpretability
Detecting Alignment Faking in Language Models
Welcome to my new site. I’ll be writing about AI safety research, alignment faking detection, and whatever else seems interesting.
SAE
Detecting Alignment Faking in Language Models
Welcome to my new site. I’ll be writing about AI safety research, alignment faking detection, and whatever else seems interesting.
multi-agent
Chaos Takes the Wheel: Salience-Weighted Attentional Hijacking via True Statements
April 2026 — bigsnarfdude
truth-contagion
Chaos Takes the Wheel: Salience-Weighted Attentional Hijacking via True Statements
April 2026 — bigsnarfdude
sae
Chaos Takes the Wheel: Salience-Weighted Attentional Hijacking via True Statements
April 2026 — bigsnarfdude
ai-safety
Chaos Takes the Wheel: Salience-Weighted Attentional Hijacking via True Statements
April 2026 — bigsnarfdude
llm-safety
Why AI Has a Split Personality (And How to Trigger the Evil Twin)
April 2026 — bigsnarfdude
prompt-injection
Why AI Has a Split Personality (And How to Trigger the Evil Twin)
April 2026 — bigsnarfdude
sycophancy
Why AI Has a Split Personality (And How to Trigger the Evil Twin)
April 2026 — bigsnarfdude
sft
DPO Does Different Things to Different Axes — Stage Attribution on talkie-1930-13b
Numbers from n=50 across base/SFT/IT on talkie-1930-13b and base/SFT/IT on OLMo-3-7B. All claims are Tier-1 (detector-level) unless explicitly noted. Recruit...
control-validation
DPO Does Different Things to Different Axes — Stage Attribution on talkie-1930-13b
Numbers from n=50 across base/SFT/IT on talkie-1930-13b and base/SFT/IT on OLMo-3-7B. All claims are Tier-1 (detector-level) unless explicitly noted. Recruit...
vpd
SVD and VPD: Two Methods for Reading What Training Wrote Into a Model’s Weights
May 6, 2026 — bigsnarfdude
weight-diff
SVD and VPD: Two Methods for Reading What Training Wrote Into a Model’s Weights
May 6, 2026 — bigsnarfdude
circuit-analysis
What Happens Inside a Language Model
The Setup