GLM 5.2 and 5.3 Autonomous Insider: autonomous AI agents misalignment evaluation
October 3, 2026 — bigsnarfdude
Posts by Tag
- llm-safety 7
- agent-evals 6
- mechanistic-interpretability 4
- forensic-concealment 4
- incident-response 4
- multi-agent 3
- ai-safety 2
- alignment 2
- stage-attribution 2
- talkie 2
- dpo 2
- developmental-interpretability 2
- evaluation-awareness 2
- alignment-faking 1
- interpretability 1
- SAE 1
- truth-contagion 1
- sae 1
- prompt-injection 1
- sycophancy 1
- sft 1
- control-validation 1
- vpd 1
- weight-diff 1
- authority-bias 1
- circuit-analysis 1
- sandbox 1
- reward-hacking 1
- rl-environments 1
- threat-model 1
- containment 1
- llm-adoption 1
- mathematics 1
- swarm-forensics 1
llm-safety
The Autonomous Insider Baker’s Dozen
September 28, 2026 — bigsnarfdude
Reward Hacking Gone Wrong: Agents That Only Get Paid for the Answer
September 24, 2026 — bigsnarfdude
A Live Swarm Leader Recruits Honest Agents into Concealment Using a Shared Message Board
September 14, 2026 · updated September 15, 2026 — bigsnarfdude
The Dead Drop Threat Model: Dormant Swarm Notes Recruit Honest Agents into Concealment
September 13, 2026 — bigsnarfdude
The Model Wipes Its Fingerprints: Unprompted Forensic Concealment in Open Weight Agents
September 4, 2026 — bigsnarfdude
Why AI Has a Split Personality (And How to Trigger the Evil Twin)
April 2026 — bigsnarfdude
agent-evals
GLM 5.2 and 5.3 Autonomous Insider: autonomous AI agents misalignment evaluation
October 3, 2026 — bigsnarfdude
The Autonomous Insider Baker’s Dozen
September 28, 2026 — bigsnarfdude
Reward Hacking Gone Wrong: Agents That Only Get Paid for the Answer
September 24, 2026 — bigsnarfdude
A Live Swarm Leader Recruits Honest Agents into Concealment Using a Shared Message Board
September 14, 2026 · updated September 15, 2026 — bigsnarfdude
The Dead Drop Threat Model: Dormant Swarm Notes Recruit Honest Agents into Concealment
September 13, 2026 — bigsnarfdude
The Model Wipes Its Fingerprints: Unprompted Forensic Concealment in Open Weight Agents
September 4, 2026 — bigsnarfdude
mechanistic-interpretability
What Happens Inside a Language Model
The Setup
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
forensic-concealment
GLM 5.2 and 5.3 Autonomous Insider: autonomous AI agents misalignment evaluation
October 3, 2026 — bigsnarfdude
A Live Swarm Leader Recruits Honest Agents into Concealment Using a Shared Message Board
September 14, 2026 · updated September 15, 2026 — bigsnarfdude
The Dead Drop Threat Model: Dormant Swarm Notes Recruit Honest Agents into Concealment
September 13, 2026 — bigsnarfdude
The Model Wipes Its Fingerprints: Unprompted Forensic Concealment in Open Weight Agents
September 4, 2026 — bigsnarfdude
incident-response
The Autonomous Insider Baker’s Dozen
September 28, 2026 — bigsnarfdude
Reward Hacking Gone Wrong: Agents That Only Get Paid for the Answer
September 24, 2026 — bigsnarfdude
A Live Swarm Leader Recruits Honest Agents into Concealment Using a Shared Message Board
September 14, 2026 · updated September 15, 2026 — bigsnarfdude
The Dead Drop Threat Model: Dormant Swarm Notes Recruit Honest Agents into Concealment
September 13, 2026 — bigsnarfdude
multi-agent
A Live Swarm Leader Recruits Honest Agents into Concealment Using a Shared Message Board
September 14, 2026 · updated September 15, 2026 — bigsnarfdude
The Dead Drop Threat Model: Dormant Swarm Notes Recruit Honest Agents into Concealment
September 13, 2026 — bigsnarfdude
Chaos Takes the Wheel: Salience-Weighted Attentional Hijacking via True Statements
April 2026 — bigsnarfdude
ai-safety
Notes from Inside Karpathy’s Funnel
October 2026 — bigsnarfdude
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...
evaluation-awareness
GLM 5.2 and 5.3 Autonomous Insider: autonomous AI agents misalignment evaluation
October 3, 2026 — bigsnarfdude
The Model Wipes Its Fingerprints: Unprompted Forensic Concealment in Open Weight Agents
September 4, 2026 — bigsnarfdude
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.
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
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
sandbox
The Model Wipes Its Fingerprints: Unprompted Forensic Concealment in Open Weight Agents
September 4, 2026 — bigsnarfdude
reward-hacking
Reward Hacking Gone Wrong: Agents That Only Get Paid for the Answer
September 24, 2026 — bigsnarfdude
rl-environments
Reward Hacking Gone Wrong: Agents That Only Get Paid for the Answer
September 24, 2026 — bigsnarfdude
threat-model
The Autonomous Insider Baker’s Dozen
September 28, 2026 — bigsnarfdude
containment
The Autonomous Insider Baker’s Dozen
September 28, 2026 — bigsnarfdude
llm-adoption
Notes from Inside Karpathy’s Funnel
October 2026 — bigsnarfdude
mathematics
Notes from Inside Karpathy’s Funnel
October 2026 — bigsnarfdude
swarm-forensics
Notes from Inside Karpathy’s Funnel
October 2026 — bigsnarfdude