giorgio/MachineSex
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

llm_society pilot v1 diagnosis + v2 knobs: raise inheritance fidelity, add elitism

Browse source 
Pilot v1: full and no_grounding indistinguishable - both arms pay a ~25%
distillation tax per generation and pin at the same mutation-selection floor,
so grounding has no differential to act on. v2 raises inheritance fidelity
(n_inherit 240->600, child epochs 2->3, founders 600x3) and adds two evenly-
applied knobs: elitism (top parent survives as an unmodified copy - overlapping
generations, compensating for lossy distillation where E11 had faithful
genotype copying) and n_parents (sharper truncation selection). v1 archived at
results/llm_society_pilot1; v2 running.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BkRLcc18rwT2Lysu6PbG7v
This commit is contained in:
Giorgio Gilestro 2026-09-07 12:38:15 +01:00
parent 16e9002773
commit 9c0210ba93
3 changed files with 34 additions and 11 deletions
Download patch file Download diff file Expand all files Collapse all files

24
src/llm/society.py
View file

@ -178,6 +178,8 @@ def run_society_experiment(cfg: dict) -> pd.DataFrame:
n_inherit = int(cfg.get("n_inherit", 240))
n_cand = int(cfg.get("n_candidates", 6))
epochs = int(cfg.get("epochs", 2))
elitism = int(cfg.get("elitism", 0))
n_parents = int(cfg.get("n_parents", max(2, N // 2)))
spec_train = int(cfg.get("spec_train", 300))
spec_epochs = int(cfg.get("spec_epochs", 2))
hard = bool(cfg.get("hard", False))
@ -226,7 +228,7 @@ def run_society_experiment(cfg: dict) -> pd.DataFrame:
fitness = np.array([fv["overall"] for fv in fit_val])
scores = s["g"] * fitness + (1.0 - s["g"]) * conf
parents = select_parents(scores, dist, max(2, N // 2), diversity=s["diversity"], lam=lam)
parents = select_parents(scores, dist, n_parents, diversity=s["diversity"], lam=lam)
# ---- rows (reporting uses the verifier in every arm; the loop does not)
for i in range(N):
@ -253,10 +255,11 @@ def run_society_experiment(cfg: dict) -> pd.DataFrame:
inherit_pool = sum([make_tasks(f, n_inherit // len(fams),
seed=seed * 104729 + t * 17 + i, hard=hard)
for i, f in enumerate(fams)], [])
n_bred = N - elitism
child_sources: list[dict] = [] # per child: parent names + weights
if s["sex"]:
from .directed import sample_merge_weights
pairs = complementary_pairs(parents, dist, N)
pairs = complementary_pairs(parents, dist, n_bred)
for c, (pa, pb) in enumerate(pairs):
w = sample_merge_weights(2, n_cand, rng)
best_i, best_v = 0, -np.inf
@ -285,7 +288,7 @@ def run_society_experiment(cfg: dict) -> pd.DataFrame:
model.delete_adapter(cname)
child_sources.append({"parents": (pa, pb), "answers": answers})
else:
for c in range(N):
for c in range(n_bred):
p = parents[c % len(parents)]
model.set_adapter(f"a{p}")
answers = generate(model, tok, [x.prompt for x in inherit_pool])
@ -293,9 +296,20 @@ def run_society_experiment(cfg: dict) -> pd.DataFrame:
del model
torch.cuda.empty_cache()
# ---- reproduce: fresh LoRA per child, trained on its source's own answers
# ---- reproduce: fresh LoRA per child, trained on its source's own answers.
# Elites (overlapping generations): the top-scoring parents survive as unmodified
# copies, applied identically in every arm — reproduction here is lossy distillation,
# so without a survivor the mutation load erases the champion each generation.
new_agents = []
for c, src in enumerate(child_sources):
for e in range(elitism):
d = root / arm / f"gen{t + 1}" / f"elite{e}"
shutil.copytree(agents[parents[e]], d, dirs_exist_ok=True)
new_agents.append(str(d))
rows.append({"experiment": name, "arm": arm, "seed": seed, "generation": t + 1,
"agent": e, "selected": False, "conformity": float("nan"),
"score": float("nan"), "metric": "parents",
"value": float(parents[e] * 100 + parents[e])})
for c, src in enumerate(child_sources, start=elitism):
data = [Task(x.family, x.prompt, _answer_of(a))
for x, a in zip(inherit_pool, src["answers"]) if _answer_of(a)]
d = root / arm / f"gen{t + 1}" / f"agent{c}"