AI Explained
Plain explanations of trending AI concepts, with live visualizations.
LLM-as-a-Verifier scales agent feedback with logit-score expectations — Verification as a scaling axis — What does it mean?
Reading the verifier's whole score distribution — not one label — turns a coarse pass/fail into a continuous score you can scale.
Discrete diffusion theory unifies denoisers, scores & bridge predictors — One object, three coordinates — What does it mean?
The three ways to train a diffusion language model — denoiser, score, and bridge — turn out to be one object written in different coordinates.
Direct-OPD transfers weak-model RL gains as log-ratio rewards — Weak-to-strong reward transfer — What does it mean?
Direct-OPD reuses a small model's before/after RL log-ratio as a dense reward to lift a stronger model — no RL on the big model.
Omnigent open-sources a meta-harness for coding agents — The agent meta-harness — What does it mean?
A meta-harness is one control layer above many coding-agent CLIs — it holds credentials, session, policy, and sandbox in one place.
LOCOS finds non-literal retrieval heads by scoring logit contribution — Logit-Contribution Scoring — What does it mean?
LOCOS finds the attention heads that answer from meaning, not copied words, by scoring how much each head writes toward the answer token.
LangChain adds dynamic subagents for code-driven orchestration — Programmatic subagent fan-out — What does it mean?
LangChain Deep Agents can now write a short script that fans out one subagent per chunk, so coverage is a property of code, not a prompt.
LACUNA tests whether LLM unlearning hits the right weights — Output-level vs weight-level unlearning evaluation — What does it mean?
LACUNA plants a fact in known weights to test whether unlearning erases it, or just hides the output while a resurfacing attack revives it.
HaloGuard ships 0.8B open constitutional safety classifier — Paired counterfactual safety data — What does it mean?
HaloGuard trains a tiny safety classifier on matched prompt pairs that flip only intent, so it learns the boundary, not the keywords.
CheckRLM corrects factual drift inside retrieval-augmented reasoning — In-chain retrieval fact-checking — What does it mean?
CheckRLM fact-checks each claim inside a reasoning model's chain of thought against retrieved evidence, then patches only the wrong step.
Google releases TabFM for zero-shot tabular prediction — Tabular in-context learning — What does it mean?
TabFM predicts on a spreadsheet it has never seen by reading the whole table as one prompt — in-context learning, no per-dataset training.
SkillCoach self-evolves rubrics to grade agentic skill-use at scale — Self-evolving rubrics — What does it mean?
SkillCoach grades how an agent uses its skills on four axes — with a rubric that rewrites its own criteria, keeping only edits that still grade trusted samples right.
kNNGuard turns LLM hidden activations into a training-free guardrail — Training-free activation-space kNN guardrail — What does it mean?
A guardrail that flags unsafe prompts by reading a frozen LLM's hidden activations and kNN-matching a 50-prompt bank — no training, up to 10× faster.
