Gahow Wang gahow
0 Followers · 0 Following
  • Joined on 2026-04-03
Repositories
19
 Projects Packages Public Activity Starred Repositories
gahow pushed to main at gahow/aituner 2026-06-30 06:10:21 +00:00
1b8f5a3af1 Integrate descriptor runtime candidates into harness
gahow pushed to main at gahow/xtrain 2026-06-30 04:38:08 +00:00
99090465bf docs: M3 — DPO results (infra correct, held-out correctness flat, over-optimization collapse)
2f827fd6d8 post-train: M3 — DPO pair-gen + training loop (verifiable arithmetic)
f3c764ce95 post-train: M3 — seq_logprob + dpo_loss autograd ops
Compare 3 commits »
gahow pushed to main at gahow/aituner 2026-06-30 04:05:05 +00:00
adb5356c4b Add advisory harness attribution and descriptor planner MVP
gahow pushed to main at gahow/xtrain 2026-06-30 04:01:11 +00:00
b39e6e7110 docs: M2a — KV-cache decode engine results (token-identical + length-dependent speedup)
eff26a0898 post-train: M2a — KV-cache incremental decode engine (token-identical)
c88e2ab88c post-train: M2 — decode primitives (rope_at + decode_attention)
Compare 3 commits »
gahow pushed to main at gahow/xtrain 2026-06-30 03:13:29 +00:00
1574e21d89 post-train: M1 — verifiable-arith eval scorer + SFT format-baseline result
gahow pushed to main at gahow/xtrain 2026-06-29 15:28:26 +00:00
cb64604496 post-train: M1 fix — enlarge arith key space + saturation guard
gahow pushed to main at gahow/xtrain 2026-06-29 15:20:54 +00:00
9c70e99ae4 post-train: M1 — verifiable arithmetic task + SFT data generator
ab32168dcc docs: post-training stack design — SFT → KV-cache → DPO → GRPO (docs/18)
Compare 2 commits »
gahow pushed to main at gahow/aituner 2026-06-29 12:42:00 +00:00
08429e5da8 Refine harness design flow overview
gahow pushed to main at gahow/aituner 2026-06-29 12:27:04 +00:00
00ba573631 Document harness design contract
gahow pushed to main at gahow/aituner 2026-06-29 11:56:53 +00:00
6ea259a0a3 Keep target topology explicit in delta projections
gahow pushed to main at gahow/aituner 2026-06-29 09:57:31 +00:00
6b4efdad82 Relax lower-frontier delta projection gate
gahow pushed to main at gahow/aituner 2026-06-29 08:22:10 +00:00
9ef9550214 Use full state for frontier projection
gahow pushed to main at gahow/xtrain 2026-06-29 08:19:19 +00:00
7a1fba95b5 docs: v12 — 1.05B long-ctx base + chat-alpha SFT quality check
fbf4ac2917 sft: assistant-only SFT (ignore-index CE) + chat-prompt greedy eval
5c27493a90 docs: backfill v9/v10 scaling runs + reframe README to v0–v10 / three phases
Compare 3 commits »
gahow pushed to main at gahow/aituner 2026-06-29 08:15:16 +00:00
8dd9ada194 Add frontier delta projection harness candidates
gahow pushed to main at gahow/aituner 2026-06-28 18:34:14 +00:00
6c84dc91d7 Document hardened topology feedback
gahow pushed to main at gahow/aituner 2026-06-28 18:28:32 +00:00
1c4ed4cab3 Document hardened harness feedback
gahow pushed to main at gahow/aituner 2026-06-28 18:00:43 +00:00
6b25d56c1f Gate GMU climb on measured improvement
gahow pushed to main at gahow/aituner 2026-06-28 17:54:03 +00:00
ee101a7c24 Harden prefill scheduler harness
gahow pushed to main at gahow/aituner 2026-06-28 17:30:36 +00:00
bfd85793f3 Prioritize uncovered prefill scheduler candidates
gahow pushed to main at gahow/aituner 2026-06-28 17:12:20 +00:00
36c301c128 Add normalized prefill scheduler harness