Build a unit full-width direction from paired, already collected activations. The positive sign is target minus control. This function performs no inference.
Usage
llm_direction(
target,
control,
context,
layer,
normalize_pairs = FALSE,
orthogonalize = FALSE,
max_bytes = 64 * 1024^2
)
# S3 method for class 'relm_direction'
print(x, ...)Arguments
- target, control
Plain double matrices with identical named pairs as rows and complete, ordered neuron names
"1"through the hidden width as columns.- context
For construction, a recorded list with
model,capture,pairs,splitsandseed; see Details and the contrast-directions vignette. For application, an independently recorded destinationmodelrecord.- layer
One whole layer: 2 through the recorded layer count for residual captures; 1 through that count for
mlp_outandattn_outcaptures.- normalize_pairs
Whether to give each pair difference unit length before averaging. Defaults to FALSE, preserving relative pair magnitudes.
- orthogonalize
Whether to remove the component parallel to the mean original control vector before final normalization. Defaults to FALSE.
- max_bytes
Whole materialization budget, 1 MiB through 512 MiB; default 64 MiB. Includes input matrices and metadata, working arrays and output.
- x
A trusted
relm_directionartifact.- ...
Must be empty.
Value
llm_direction() returns a relm_direction data frame with integer
neuron, double value and a versioned direction metadata attribute.
No model pointer, input matrix or trace is retained.
Details
Context records a model checksum, architecture, quantization, hidden width,
layer count and patched engine revision. Capture is full-width residual,
mlp_out or attn_out space at
the last prompt position, raw text, embedded tokenizer, add_special = TRUE,
parse_special = FALSE, no external template, plus resolved backend, context
length and package version. Pairs record ordered IDs, exact UTF-8 prompt
checksums and source positions. The split manifest declares disjoint,
nonempty construction, selection and evaluation sets before fitting.
At most 4096 pairs, 65536 coordinates, 16384 split rows and 8 MiB of context
are admitted. Computation is row-wise; a whole difference matrix is never
materialized. Zero, nearly cancelled or nonfinite directions are errors, not
silently dropped observations. Norm guards use 64 * .Machine$double.eps.
Trusted saveRDS() and readRDS() preserve the artifact. Application and
printing validate schema, unit values, coordinates and canonical checksums.
Checksums use one bounded temporary file at a time, removed even on error.
They detect modification but do not authenticate recorded provenance or
already-loaded weights. Caller deserialization and native model/context
memory are outside this budget. Never read untrusted RDS files.
Residual artifacts retain schema 1. Component artifacts use schema 2, with
component-specific identity and a versioned prefix in every checksum domain.
Construction does not prove that a model supports editing the recorded site.