The feed
Published findings
Every change the detector flagged against a model's own witnessed baseline. Pass-rate Movements survive multiple-comparison correction; thinking-token Movements cross a relative-shift threshold. Watches are unconfirmed; latency is operational context.
sarvam-105b@sarvam · Sycophancy probes pass rate fellWatch 2026-11-09
deepseek-v4-flash-0731@deepinfra · Median latency roseContext 2026-11-09
grok-4.7@xai · Median latency roseContext 2026-11-09
inkling-small@deepinfra · Median latency roseContext 2026-11-09
llama-4-maverick@deepinfra · Median latency roseContext 2026-11-09
muse-spark-1.3@meta · Median latency roseContext 2026-11-09
nemotron-3-ultra@deepinfra · Median latency roseContext 2026-11-09
hy4-preview@deepinfra · Thinking-token use rose on the instruction following probe setWatch 2026-11-09
mistral-large-4-0@mistral · Thinking-token use rose on the instruction following probe setMovement 2026-11-08
glm-5.2@deepinfra · Instruction following pass rate fellWatch 2026-11-08
glm-5.2@deepinfra · Median latency roseContext 2026-11-08
minimax-m2.5@deepinfra · Median latency fellContext 2026-11-08
mimo-v2.6-flash@deepinfra · Thinking-token use rose on the instruction following probe setMovement 2026-10-31
minimax-m3.1-flash@minimax · Thinking-token use rose on the instruction following probe setMovement 2026-10-30
deepseek-reasoner@deepseek · Tool call pass rate fellMovement 2026-10-25
hy4-preview@deepinfra · Thinking-token use rose on the instruction following probe setMovement 2026-10-21
hy4-preview@deepinfra · Thinking-token use rose on the sycophancy probe setMovement 2026-10-21
deepseek-chat@deepseek · Instruction following pass rate fellMovement 2026-10-17
hy4-preview@deepinfra · Thinking-token use rose on the verbosity probe setMovement 2026-10-13
qwen3.8-max@deepinfra · Thinking-token use rose on the sycophancy probe setMovement 2026-10-12
gpt-6-luna · Thinking-token use rose on the instruction following probe setMovement 2026-10-03
gpt-6-luna · Thinking-token use rose on the sycophancy probe setMovement 2026-10-03
gpt-6-luna · Thinking-token use fell on the structured output probe setMovement 2026-10-02