Siyuc commited on
Commit
148706a
·
verified ·
1 Parent(s): 4f45eac

Add HRM KVReuse12 step114274 evaluation answers and 29 scored benchmarks

Browse files
evaluations/hrm-text/20261008-hrm-kvreuse12-step114274/README.md ADDED
@@ -0,0 +1,56 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # HRM KVReuse12 EMA: step 114274
2
+
3
+ All 112,420 answers generated; 29/31 benchmarks scored. HumanEval+ and MBPP+ pass@1 remain pending isolated EvalPlus execution.
4
+
5
+ ## hrm-text-official
6
+
7
+ | Benchmark | Metric | Score (%) |
8
+ |---|---|---:|
9
+ | GSM8k | acc | 45.26 |
10
+ | MATH | acc | 27.70 |
11
+ | DROP | f1 | 27.96 |
12
+ | MMLU | acc | 36.57 |
13
+ | ARC | acc | 33.70 |
14
+ | HellaSwag | acc | 30.09 |
15
+ | Winogrande | acc | 50.51 |
16
+ | BoolQ | acc | 38.00 |
17
+ | BoolQ | macro_f1 | 27.77 |
18
+
19
+ ## hrm-text-extended-v1
20
+
21
+ | Benchmark | Metric | Score (%) |
22
+ |---|---|---:|
23
+ | triviaqa | canonical_em | 2.02 |
24
+ | triviaqa | canonical_f1 | 3.80 |
25
+ | nq | canonical_em | 1.77 |
26
+ | nq | canonical_f1 | 4.99 |
27
+ | webq | canonical_em | 8.32 |
28
+ | webq | canonical_f1 | 16.65 |
29
+ | webq | all_answer_em | 13.44 |
30
+ | webq | all_answer_f1 | 22.79 |
31
+ | lambada_openai | acc | 20.36 |
32
+ | tydiqa_goldp | canonical_em | 48.86 |
33
+ | tydiqa_goldp | canonical_f1 | 61.92 |
34
+ | squadv2 | canonical_em | 51.18 |
35
+ | squadv2 | canonical_f1 | 66.74 |
36
+ | coqa | canonical_em | 14.03 |
37
+ | coqa | canonical_f1 | 15.90 |
38
+ | svamp | acc | 58.00 |
39
+ | asdiv | acc | 64.69 |
40
+ | mawps | acc | 79.18 |
41
+ | induction_head | acc | 0.40 |
42
+ | var_assign_d0_math | acc | 97.70 |
43
+ | var_assign_d0_code | acc | 98.70 |
44
+ | var_assign_d1_math | acc | 36.20 |
45
+ | var_assign_d1_code | acc | 27.80 |
46
+ | bigbench_dyck_languages | acc | 0.60 |
47
+ | bigbench_qa_wikidata | canonical_em | 33.80 |
48
+ | bigbench_qa_wikidata | canonical_f1 | 34.21 |
49
+ | arc_easy | acc | 44.78 |
50
+ | bigbench_cs_algorithms | acc | 6.52 |
51
+ | bigbench_cs_algorithms | strict_acc | 5.45 |
52
+ | bbh_cot | acc | 8.72 |
53
+ | math500 | acc | 25.00 |
54
+ | humaneval_plus | pass@1 | pending |
55
+ | mbpp_plus | pass@1 | pending |
56
+
evaluations/hrm-text/20261008-hrm-kvreuse12-step114274/extended-hrm_kvreuse12_5e191387.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33316bef22382cf63c6a07232d218f200be192cdc254de1dfe4539e8cc1a5f97
3
+ size 6544322
evaluations/hrm-text/20261008-hrm-kvreuse12-step114274/manifest.json ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "snapshot": "20261008-hrm-kvreuse12-step114274",
3
+ "checkpoint": "small_search_1epoch/HRM-H4-L3-W768-H11-L8-BP16-KVReuse12-causal-AdamATan2-EMA-seed0-1epoch__20261007-031120_hrm-h4-bp16-kvreuse12-h20x16--5e191387/step_114274",
4
+ "weights": "ema",
5
+ "scored_benchmarks": 29,
6
+ "generated_benchmarks": 31,
7
+ "examples": 112420,
8
+ "files": [
9
+ {
10
+ "path": "README.md",
11
+ "bytes": 1665,
12
+ "sha256": "681f8645136f31cb0e13fa03bffff78855ca8bbfd03eb708b8cde8d0a5c83d2c"
13
+ },
14
+ {
15
+ "path": "extended-hrm_kvreuse12_5e191387.tar.gz",
16
+ "bytes": 6544322,
17
+ "sha256": "33316bef22382cf63c6a07232d218f200be192cdc254de1dfe4539e8cc1a5f97"
18
+ },
19
+ {
20
+ "path": "official-hrm_kvreuse12_5e191387.tar.gz",
21
+ "bytes": 3778328,
22
+ "sha256": "7d0024d67c62444a7696db7fbe008761cc64b8dc6216f7487780163fbdc27b04"
23
+ }
24
+ ]
25
+ }
evaluations/hrm-text/20261008-hrm-kvreuse12-step114274/official-hrm_kvreuse12_5e191387.tar.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d0024d67c62444a7696db7fbe008761cc64b8dc6216f7487780163fbdc27b04
3
+ size 3778328