Reference

Quick tables. variants() and tasks() are the live source of truth.

Node presets

symbol	family	status
:falandays_base (:falandays)	Falandays	stable baseline (authors-faithful)
:falandays_extended	Falandays	base + noise + Watts–Strogatz + Dale
:falandays_noisy / _ablated / _hemispheric / _oosawa / _dendritic / _spatial / _delayed	Falandays	experimental variants
:compartmental_dense / _structured	CTRNN	evolved (required)
:sorn	SORN	experimental
:null_random	control	random-output null baseline (sanity floor)

Tasks

:wall, :tracking, :pong, :pong_hitrate, :cartpole (+ _hard / _swingup / _long), :torus, :forage.

Ablations

:freeze_plasticity, :zero_recurrent, :clamp_target, :disable_vision (cross-family perturbations), plus the compartmental :reset_dendrites, :no_soma_back, :no_hillock_back.

Sweep config schema

[sweep]     # id, mode = one_at_a_time | factorial, seeds = [...], max_cells
[baseline]  # node, task, N, ticks — the canonical setup (see below)
[axes]      # "node.<p>" / "env.<p>" / "drive.<p>" / "task.<p>" / "ablation" / "seed" -> [values]
[analytics] # measures = [...] — see the full set below

See Tooling for the walkthrough and sweep --list-axes <node> <task> to discover what’s tunable.

[baseline] keys — the canonical setup

The [baseline] table fixes the setup that every axis then perturbs. It accepts a small set of canonical scalar keys plus any dotted axis key; the dotted keys and the env.* / node.* / drive.* knobs pass straight through to the corresponding simulate(...) kwargs (so anything a run accepts, a baseline can pin).

Canonical scalar keys:

key	meaning
node	node preset symbol (default falandays_base)
task	task symbol (default wall)
N (alias n_nodes)	reservoir size
ticks	rollout length
window	metric / liveness window (defaults to ticks, or the task default for swarm tasks)
n_agents	swarm population size (swarm/forage tasks)
seed_base (alias seed)	base seed
body / drive	body and drive preset symbols
ablation	ablation symbol applied to the baseline

Common dotted pass-through knobs (forage / swarm):

key	meaning
"env.vision_range"	max conspecific/source vision distance
"env.source_gain"	forage source bank gain
"env.conspecific_vision"	whether agents see each other (true / false)
"env.capture_radius"	forage capture radius
"env.sensory_noise"	sensory noise level

[analytics] measures — the full set

measures selects which analytics are computed per cell (default ["sigma_mr", "spectral_radius", "liveness"]). Valid names:

measure	what it records
sigma_mr	pooled branching ratio (MR)
sigma_mr_node	per-agent reservoir branching
sigma_mr_agent	agent-scale turn-event branching (per ensemble observable)
dist_to_source	mean distance to forage source
forage_score	foraging performance score
spectral_radius	dominant recurrent-eigenvalue magnitude
susceptibility_node	node-scale synchrony susceptibility
susceptibility_agent	agent-scale polarization susceptibility
correlation_length	velocity-fluctuation correlation length
contact_clusters	contact-graph connected-component summaries
liveness	run liveness (already in the base columns)
regime	discrete swarm-regime label (+ polarization / milling / speed)

contact_clusters emits cluster_n_components, cluster_largest_component_frac, and cluster_mean_component_size; regime emits regime_polarization, regime_milling, and regime_speed alongside regime. The aliases clusters, cluster_stats, and contact_graph_clusters all canonicalize to contact_clusters.