TECHNICAL DICTIONARY · 60 CANONICAL TERMS · FULL DOMAIN INDEX
Field Guide Glossary
Curated in pedagogical sequence: from fundamental parameter geometry to transformer circuit mechanics, empirical verification criteria, and benchmark protocols. Every entry directly anchors to its corresponding guide chapter.
/focus
70/ 70
Quick terms:
Initial:
- Parameter Ch 1
- One number inside a network - a weight or a bias. The knobs.
- Weight Ch 1
- A parameter that multiplies an input (or another neuron’s output) on its way through the network.
- Bias Ch 1
- A parameter added after the multiplying, shifting a neuron’s trigger point.
- θ (theta) Ch 1
- The vector of ALL parameters, treated as one long list of numbers.
- P (parameter count) Ch 1
- How many numbers θ contains - 17,253 for the Exp-1 net, ~0.8 billion for Qwen3.5-0.8B.
- Architecture Ch 1
- The wiring diagram: how many layers, of what kind, connected how. Same architecture + different weights = different skill.
- Weight space Ch 1
- The space of every possible setting of θ - training walks through it, edits jump across it.
- MLP Ch 1
- Multi-layer perceptron: the simplest network - layers of matrix-multiply + nonlinearity.
- tanh / ReLU Ch 1
- Nonlinear functions applied after each layer; tanh squashes to (−1,1), ReLU keeps positives, zeroes negatives.
- Receptive field Ch 4
- The slice of input a given weight column responds to - why a 10×10 grid can reveal a paddle row.
- Forward pass Ch 4
- Running the network on an input to get an output. Weight reading in the READ experiments happened with zero forward passes.
- Gradient descent Ch 2
- The default weight-writer: nudge every parameter a tiny step downhill on the loss, millions of times.
- SGD / Adam / Muon Ch 2
- Flavours of gradient descent - different recipes for shaping the tiny steps.
- Learning rate (η) Ch 2
- The size of each training step. Too big: overshoot. Too small: forever.
- Loss function Ch 2
- A number measuring how wrong the network is - the hill training descends.
- PPO Ch 6
- Proximal Policy Optimization - the RL algorithm that trained the paper’s game baselines.
- Hypernetwork Ch 2
- A network trained to output another network’s weights (Ha et al., 2016).
- Fast-weight programmer Ch 2
- A transformer variant that writes temporary weights as it runs (Schlag et al., 2021).
- Tracr Ch 2
- A compiler from formal programs (RASP) into toy-transformer weights - the third way to write weights.
- Weight-space programming (WSP) Ch 2
- The paper’s claim: a general-purpose LLM reads θ, reasons, and writes Δ directly - no gradients, optionally a few forward passes for feedback.
- Δ (delta) Ch 2
- The edit: the vector added to θ. In Exp 8 it touched 131,072 of ~0.8B weights.
- Fine-tuning Ch 3
- Training an existing model further on new data - gradients again, shorter run.
- LoRA Ch 3
- Low-Rank Adaptation: learn two thin matrices whose product is the change; add it to frozen weights.
- Task arithmetic Ch 3
- Fine-tune, subtract the base: the difference is a “task vector” you can add or remove (Ilharco et al., 2022).
- Model merging Ch 3
- Combining several trained models’ weights into one (TIES and friends).
- Model editing Ch 3
- Surgical weight changes targeting one specific fact or behaviour (ROME, Meng et al. 2022).
- Activation steering Ch 3
- Nudging the model’s inner state at run time without touching weights.
- Mechanistic interpretability Ch 3
- Reverse-engineering networks into human-readable circuits - the detective cousin, and the source of the copying-head technique.
- Weight-space learning Ch 3
- The research field covering understanding, representing, and generating weights (survey: arXiv 2603.10090 [VERIFY]).
- Transformer Ch 4
- The architecture behind modern language models: attention + MLPs in a residual stream.
- Attention head Ch 4
- One pattern-matching unit inside attention; heads numbered layer.head (e.g. 3.1).
- Causal mask Ch 4
- The rule that each token can attend only to earlier tokens - the lower triangle in the attention grid.
- Induction head Ch 4
- A two-head circuit implementing [A][B]…[A]→B - the engine of in-context copying.
- Copying head Ch 4
- The heads Exp 3 identified from weights alone, matching measured ground truth exactly.
- Composition score Ch 4
- How much two heads’ circuits multiply together - the weight-only signal that finds copiers.
- Residual stream Ch 5
- The shared “conveyor belt” a transformer adds information into at every layer.
- DDPM Ch 4
- Denoising diffusion: learn to remove noise step by step; the tiny image-maker of Exp 2.
- Sycophancy Ch 4
- Agreeing with the user instead of being honest - the behaviour Exp 7 tried to detect from weights.
- Kinetix Ch 5
- The physics environment whose ramp the actor couldn’t climb (Exp 4).
- MinAtar Ch 6
- Tiny Atari-inspired games (Breakout, Seaquest, Asterix) used for the WRITE experiments.
- Go (9×9) Ch 6
- The small Go board on which a written-from-scratch policy beat a PPO trained on 20M games.
- Liberty Ch 7
- In Go, an empty point adjacent to a stone - counting them was a stated ingredient of the written policy.
- Atari Ch 7
- In Go, a group with one liberty left - urgent. (Also: the game company; context tells.)
- Direction (in weight space) Ch 5
- A vector u along which behaviour can be added or removed; refusal and thought-loops are direction-mediated.
- Rank-1 projection Ch 5
- The operation W′ = (I − uuᵀ)W that erases direction u from a weight matrix.
- Thought loop / overthinking Ch 5
- The self-reinforcing pattern that made Qwen3.5-0.8B hit the token limit on 93% of math problems.
- MATH-500 / GSM8K / IFEval Ch 5
- Benchmarks: competition maths / grade-school word problems / instruction following.
- Negative control Ch 5
- A run that SHOULD show no effect (Gemma 3 270M on math) - its flatness validates the method.
- Edit sparsity (s) Ch 5
- The fraction of parameters an edit touches: s = ‖Δ‖₀/P. Exp 8: 0.016%.
- Blast radius Ch 5
- Informal name for sparsity - how far the edit’s fingerprint reaches.
- Functional drift (D) Ch 12
- How much the edit changed outputs on held-out inputs (KL divergence) - the containment metric WSP-Bench-Lite demands.
- Budget curve (fidelity(n)) Ch 5
- Quality vs allowed forward passes n - reported points only, never smooth interpolation.
- Non-monotonic Ch 8
- Goes up, then down, then up: the 25-pass IFEval dip (49.0 → 39.9). A signature of n=1 noise.
- n=1 run Ch 8
- A single execution per condition - the standing caveat on every number in the paper.
- Baseline Ch 8
- The rival method that must be tried before claiming novelty - decoding tweaks, standard direction ablation, plain code. The review’s missing piece.
- Truth contract Ch 13
- This guide’s rule: every claim wears exactly one badge - 🟦 paper-reported, 🟩 reviewed, 🟨 schematic, 🟥 speculation.
- WSP-Bench-Lite Ch 12
- Het’s proposed community benchmark: Read/Fix/Write tracks, ≥5 seeds, mandatory baselines, author-blind tasks, public logs.
- Permutation symmetry Ch 1
- Hidden units can be relabelled without changing the network’s function - the reason naive weight-reading is underdetermined.
- Weight fingerprint Ch 4
- A characteristic spatial pattern in first-layer weight norms (e.g. 58% of agent norm on row 8 in Breakout) that reveals a network’s task or environment without running forward passes.
- Direction ablation Ch 5
- Projecting out a specific 1D subspace u from model weights or residual activations: W′ = (I − uuᵀ)W to excise a target behaviour.
- Abliteration Ch 5
- The open-source community name for removing refusal or looping directions from open-weight models via weight projection without fine-tuning or gradients.
- Forward-pass budget Ch 6
- The permitted number of forward evaluation calls (n ∈ {0, 10, 25, 50}) allowed for candidate verification during weight synthesis or repair.
- Neural Functional Network (NFN) Ch 3
- A meta-neural architecture designed to process neural network weights directly while respecting permutation symmetries across hidden layers (Navon et al., 2023).
- Git Re-Basin Ch 3
- The discovery that two neural networks trained independently from different seeds can be aligned by permutation to achieve zero barrier on the linear interpolation path: L(αθ_A + (1-α)Πθ_B) ≈ 0 (Ainsworth et al., 2022).
- DARE (Drop And REscale) Ch 3
- A sparsification algorithm that randomly drops up to 90-99% of fine-tuned delta weights and rescales the remainder, enabling interference-free multi-model merging (Yu et al., 2024).
- Model Stitching Ch 3
- Connecting early layers of model A to late layers of model B with an affine stitch layer to test representation compatibility across architectures (Lenc & Vedaldi, 2015).
- Weight Diffusion Ch 2
- Training a generative diffusion model to sample valid, functioning neural network parameter tensors from Gaussian noise (Peebles et al., 2023).
- Intrinsic Dimension Ch 1
- The minimal number of degrees of freedom required to solve an objective in weight space, showing that deep networks operate in very low-dimensional active subspaces (Li et al., 2018).
Guide canonical · Ch 1
Guide canonical · Ch 1
Guide canonical · Ch 1
Guide canonical · Ch 1
Guide canonical · Ch 1
Guide canonical · Ch 1
Guide canonical · Ch 1
Guide canonical · Ch 1
Guide canonical · Ch 1
Guide canonical · Ch 4
Guide canonical · Ch 4
Guide canonical · Ch 2
Guide canonical · Ch 2
Guide canonical · Ch 2
Guide canonical · Ch 2
Guide canonical · Ch 6
Guide canonical · Ch 2
Guide canonical · Ch 2
Guide canonical · Ch 2
Guide canonical · Ch 2
Guide canonical · Ch 2
Guide canonical · Ch 3
Guide canonical · Ch 3
Guide canonical · Ch 3
Guide canonical · Ch 3
Guide canonical · Ch 3
Guide canonical · Ch 3
Guide canonical · Ch 3
Guide canonical · Ch 3
Guide canonical · Ch 4
Guide canonical · Ch 4
Guide canonical · Ch 4
Guide canonical · Ch 4
Guide canonical · Ch 4
Guide canonical · Ch 4
Guide canonical · Ch 5
Guide canonical · Ch 4
Guide canonical · Ch 4
Guide canonical · Ch 5
Guide canonical · Ch 6
Guide canonical · Ch 6
Guide canonical · Ch 7
Guide canonical · Ch 7
Guide canonical · Ch 5
Guide canonical · Ch 5
Guide canonical · Ch 5
Guide canonical · Ch 5
Guide canonical · Ch 5
Guide canonical · Ch 5
Guide canonical · Ch 5
Guide canonical · Ch 12
Guide canonical · Ch 5
Guide canonical · Ch 8
Guide canonical · Ch 8
Guide canonical · Ch 8
Guide canonical · Ch 13
Guide canonical · Ch 12
Guide canonical · Ch 1
Guide canonical · Ch 4
Guide canonical · Ch 5
Guide canonical · Ch 5
Guide canonical · Ch 6
Guide canonical · Ch 3
Guide canonical · Ch 3
Guide canonical · Ch 3
Guide canonical · Ch 3
Guide canonical · Ch 2
Guide canonical · Ch 1
Need full formal literature citations?