dataset_name stringlengths 45 72 | script_name stringclasses 10
values | model stringclasses 4
values | hyperparameters stringclasses 5
values | input_datasets stringclasses 2
values | description stringlengths 212 1.03k | tags stringlengths 66 97 | custom_metadata stringlengths 139 147 | updated stringlengths 32 32 | experiment_id null | run_id null | artifact_type null | visualizer_type null | artifact_group null | parent_artifact null | size_bytes int64 -1 -1 | created stringlengths 32 32 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
browsecomp-ctxgraph-30b-rl-fusedent-canary-v1 | train_bc_ctxgraph_30b_instruct_9node_2h_fusedent_canary_yw.sh | Qwen3-30B-A3B-Instruct-2507 | {"lr": "2e-6", "kl_coef": 0.005, "kl_loss_coef": 0.0005, "entropy_coeff": 0.005, "use_fused_kernels": true, "fused_backend": "torch", "max_turn": 100, "max_session": 10, "rollout_quant": "fp8", "gpu_memory_utilization": 0.55, "train_batch_size": 14, "rollout_n": 8, "response_length": 32768, "failure_shaping": 0.2, "jud... | [] | Fused-kernels entropy canary (job vista:820770, 2026-07-10). First run ever with a working entropy bonus: entropy_coeff=0.005 via use_fused_kernels=True (FusedLinearForPPO). Step-1 actor update survived (787422 OOMed here). actor/entropy=0.346, actor/entropy_loss=0.00173, step time 50.5 min, mem peak 101.4GB, val_befor... | ["browsecomp-ctxgraph-30b-rl", "canary", "entropy-regularization"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:820770", "cluster": "vista", "artifact_status": "final", "canary": true} | 2026-07-11T20:07:29.354041+00:00 | null | null | null | null | null | null | -1 | 2026-07-11T20:07:29.354041+00:00 |
browsecomp-ctxgraph-30b-rl-fusedent-prod-metrics-v1 | train_bc_ctxgraph_30b_instruct_32node_8h_fusedent_prod_yw.sh | Qwen3-30B-A3B-Instruct-2507 | {"lr": "2e-6", "kl_coef": 0.005, "kl_loss_coef": 0.0005, "entropy_coeff": 0.005, "use_fused_kernels": true, "fused_backend": "torch", "max_turn": 100, "max_session": 10, "rollout_quant": "fp8", "gpu_memory_utilization": 0.55, "train_batch_size": 14, "rollout_n": 8, "response_length": 32768, "failure_shaping": 0.2, "jud... | [] | Production run 1/3 (job vista:823129, 2026-07-11/12, 8h TIMEOUT by design, steps 1-15 of 40). First 30B run with working entropy regularization at 28-trainer scale. Val task_reward: 0.527(s0) 0.553(s3) 0.520(s6) 0.520(s9) 0.553(s12) 0.613(s15) — step-15 val 0.613 is the ALL-TIME HIGH across all 30B runs (prev best fait... | ["browsecomp-ctxgraph-30b-rl", "production", "entropy-regularization"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:823129", "cluster": "vista", "artifact_status": "partial", "canary": false} | 2026-07-12T19:37:52.599642+00:00 | null | null | null | null | null | null | -1 | 2026-07-12T19:37:52.599642+00:00 |
browsecomp-ctxgraph-30b-rl-fusedent-prod-metrics-v2 | train_bc_ctxgraph_30b_instruct_32node_8h_fusedent_prod_yw.sh | Qwen3-30B-A3B-Instruct-2507 | {"lr": "2e-6", "kl_coef": 0.005, "kl_loss_coef": 0.0005, "entropy_coeff": 0.005, "use_fused_kernels": true, "fused_backend": "torch", "max_turn": 100, "max_session": 10, "rollout_quant": "fp8", "gpu_memory_utilization": 0.55, "train_batch_size": 14, "rollout_n": 8, "response_length": 32768, "failure_shaping": 0.2, "jud... | [] | Production run 2/3 (job vista:825290, 2026-07-12/13, 8h TIMEOUT by design, resumed from global_step_15, steps 16-31 of 40). Val task_reward: 0.567(s15 redo) 0.567(s18) 0.580(s21) 0.553(s24) 0.580(s27) 0.600(s30). Combined with run 1: SEVEN consecutive vals >=0.55 (s12 through s30) — pre-registered success bar (3 consec... | ["browsecomp-ctxgraph-30b-rl", "production", "entropy-regularization"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:825290", "cluster": "vista", "artifact_status": "partial", "canary": false} | 2026-07-13T03:52:32.661220+00:00 | null | null | null | null | null | null | -1 | 2026-07-13T03:52:32.661220+00:00 |
browsecomp-ctxgraph-30b-rl-fusedent-prod-metrics-v3 | train_bc_ctxgraph_30b_instruct_32node_8h_fusedent_prod_yw.sh | Qwen3-30B-A3B-Instruct-2507 | {"lr": "2e-6", "kl_coef": 0.005, "kl_loss_coef": 0.0005, "entropy_coeff": 0.005, "use_fused_kernels": true, "fused_backend": "torch", "max_turn": 100, "max_session": 10, "rollout_quant": "fp8", "gpu_memory_utilization": 0.55, "train_batch_size": 14, "rollout_n": 8, "response_length": 32768, "failure_shaping": 0.2, "jud... | [] | Production run 3/3 FINAL (job vista:826755, 2026-07-13, COMPLETED exit 0, steps 32-40 of 40 — FULL RUN DONE). Vals this job: 0.640(s31 redo, all-time high reading) 0.587(s33) 0.527(s36) 0.593(s39) 0.633(s40 final). Complete 40-step trajectory: baseline 0.527 -> final 0.633, peak readings 0.613/0.640. 10 of 11 vals from... | ["browsecomp-ctxgraph-30b-rl", "production", "entropy-regularization"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:826755", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-07-13T17:19:30.596948+00:00 | null | null | null | null | null | null | -1 | 2026-07-13T17:19:30.596948+00:00 |
browsecomp-ctxgraph-30b-rl-fold-fusedent-prod-metrics-v1 | train_bc_foldagent_30b_instruct_32node_8h_fusedent_prod_yw.sh | Qwen3-30B-A3B-Instruct-2507 | {"lr": "2e-6", "kl_coef": 0.005, "kl_loss_coef": 0.0005, "entropy_coeff": 0.005, "use_fused_kernels": true, "fused_backend": "torch", "max_turn": 100, "max_session": 10, "rollout_quant": "fp8", "gpu_memory_utilization": 0.55, "train_batch_size": 14, "rollout_n": 8, "response_length": 32768, "failure_shaping": 0.2, "jud... | [] | FoldAgent COMPARISON ARM run 1 (job vista:826804, 2026-07-13, 8h TIMEOUT, steps 1-13 of 40). Same optimization envelope as ctxgraph arm (entropy_coeff=0.005 + fused kernels, LR=2e-6, KL=0.005/0.0005, max_turn/max_session=100/10, FP8, gpu_mem=0.55), fold_agent loop with fold-standard reward ([flat,scope], lambda_cost=0.... | ["browsecomp-ctxgraph-30b-rl", "fold-comparison-arm", "entropy-regularization"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:826804", "cluster": "vista", "artifact_status": "partial", "canary": false} | 2026-07-13T17:19:40.217835+00:00 | null | null | null | null | null | null | -1 | 2026-07-13T17:19:40.217835+00:00 |
browsecomp-ctxgraph-30b-rl-cleanrun-ctxgraph-v1 | train_bc_ctxgraph_30b_instruct_32node_8h_fusedent_prod_yw.sh | Qwen3-30B-A3B-Instruct-2507 | {"lr": "2e-6", "kl_coef": 0.005, "kl_loss_coef": 0.0005, "entropy_coeff": 0.005, "use_fused_kernels": true, "fused_backend": "torch", "max_turn": 100, "max_session": 10, "rollout_quant": "fp8", "gpu_memory_utilization": 0.55, "train_batch_size": 14, "rollout_n": 8, "response_length": 32768, "failure_shaping": 0.2, "jud... | [] | CLEAN-RECIPE ctxgraph run 1 (job vista:834557, 2026-07-15/16, 8h TIMEOUT, steps 1-16 of 30). All audit fixes active: fractional token-level graph rewards (first time ever), strict judge (no relaxed_em refusal channel), greedy val. Clean-scale vals: 0.440(s0 baseline) 0.393(s3) 0.400(s6) 0.427(s9) 0.467(s12) 0.453(s15) ... | ["browsecomp-ctxgraph-30b-rl", "production", "entropy-regularization"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:834557", "cluster": "vista", "artifact_status": "partial", "canary": false} | 2026-07-16T06:12:19.274623+00:00 | null | null | null | null | null | null | -1 | 2026-07-16T06:12:19.274623+00:00 |
browsecomp-ctxgraph-30b-rl-tunedexp-ctxgraph-v1 | train_bc_ctxgraph_30b_instruct_32node_8h_tunedexp_prod_yw.sh | Qwen3-30B-A3B-Instruct-2507 | {"lr": "2e-6", "kl_coef": 0.005, "kl_loss_coef": 0.0005, "entropy_coeff": 0.005, "use_fused_kernels": true, "fused_backend": "torch", "max_turn": 100, "max_session": 10, "rollout_quant": "fp8", "gpu_memory_utilization": 0.55, "train_batch_size": 7, "rollout_n": 16, "response_length": 32768, "branch_len": 16384, "consol... | [] | TUNED+EXPLORATION ctxgraph arm, FINAL (runs vista:837826 steps 0-16 + vista:839112 steps 17-33, cancelled at s33 by user decision 2026-07-18 — depth verdict negative). Recipe: tuned reward (de-gated fail shaping, uniqueness x10, anti-farm structural) + deep-groups exploration (rollout n=16, batch=7) + branch_len 16384 ... | ["browsecomp-ctxgraph-30b-rl", "production", "deep-groups", "tuned-reward"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:837826+839112", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-07-19T02:42:20.152040+00:00 | null | null | null | null | null | null | -1 | 2026-07-17T15:11:19.903570+00:00 |
browsecomp-ctxgraph-30b-rl-noniso-clean-k20-v1 | train_bc_ctxgraph_30b_instruct_32node_8h_noniso_clean_k20_yw.sh | Qwen3-30B-A3B-Instruct-2507 | {"lr": "2e-6", "kl_coef": 0.005, "kl_loss_coef": 0.0005, "entropy_coeff": 0.005, "use_fused_kernels": true, "agent_loop": "context_graph_agent (non-isolated)", "search_show_topk": 20, "max_turn": 100, "max_session": 10, "rollout_quant": "fp8", "gpu_memory_utilization": 0.55, "ppo_max_token_len_per_gpu": 40960, "train_b... | [] | FIRST non-isolated ctxgraph RL training run (job vista:860532, 2026-07-23/24, 8h TIMEOUT, 15 steps, ckpt 15, ZERO EngineCore crashes — reserved stable ~89GB, first machine-stable non-iso training after EngineCore root-cause fix confirmed: crashes were GPU reserved exceeding 96GB HBM, not host OOM). Config: context_grap... | ["browsecomp-ctxgraph-30b-rl", "non-isolated", "k20-retrieval", "training"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:860532", "cluster": "vista", "artifact_status": "partial", "canary": false} | 2026-07-24T17:17:08.280203+00:00 | null | null | null | null | null | null | -1 | 2026-07-24T17:17:08.280203+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-239q-v1 | rejudge_db_azure_yw.py (post-hoc judging of eval_discoverybench_qwen3_30b_instruct_8node.sh outputs) | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench real-test formal eval, ctxgraph, 239 queries, max_turn=24. 146/239 tasks produced discovery_result.json (rows here); 93 produced none (score 0). Judged post-hoc with Azure gpt-5-nano (same judge model + prompts as Huancheng's baselines). Mean HMS: 0.1197 over 146 answered, 0.0731 over all 239. | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph", "qwen3-30b"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:926252", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-21T15:46:37.615314+00:00 | null | null | null | null | null | null | -1 | 2026-08-21T15:46:37.615314+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-react-239q-v1 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge_db_azure_yw.py backfill of rate-limited judgings) | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench real-test formal baseline, ReAct (no memory structure), 239 queries, max_turn=24, judge gpt-5-nano (Azure; same prompts/model as ctxgraph run). 182/239 answered; mean HMS 0.0997 answered / 0.0759 strict-239; 99 nonzero. Elapsed 26m19s x 8 nodes = 3.5 node-hours. Tasks with no discovery_result.json have n... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "react", "qwen3-30b"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:927276", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-21T17:19:37.063299+00:00 | null | null | null | null | null | null | -1 | 2026-08-21T17:19:37.063299+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-fold-239q-v1 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge_db_azure_yw.py backfill of rate-limited judgings) | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench real-test formal baseline, FoldAgent (linear folding memory), 239 queries, max_turn=24, judge gpt-5-nano (Azure; same prompts/model as ctxgraph run). 188/239 answered; mean HMS 0.1038 answered / 0.0817 strict-239; 109 nonzero. Elapsed 37m23s x 8 nodes = 5.0 node-hours. Tasks with no discovery_result.json... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "fold", "qwen3-30b"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:927277", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-21T17:19:46.425830+00:00 | null | null | null | null | null | null | -1 | 2026-08-21T17:19:46.425830+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-discipline-239q-v1 | eval_discoverybench_qwen3_30b_instruct_8node.sh (in-rollout Azure judging; prompt patch in agents/graph_agent_code_isolated.py) | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph RERUN with output-discipline prompt (write draft discovery_result.json early + short answer-writing code blocks; ctxgraph-only patch in graph_agent_code_isolated.py). 157/239 answered (was 146 unpatched), judged in-rollout by Azure gpt-5-nano, 0 judge errors. Mean HMS 0.1025 answered / 0.0673 st... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph", "qwen3-30b"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:927518", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-21T20:22:35.167682+00:00 | null | null | null | null | null | null | -1 | 2026-08-21T20:22:35.167682+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-react-239q-v2 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench react repeat 2 of 3, vista job 927665. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 177/239 answered, mean HMS 0.0839 over answered / 0.0622 strict-239. Part of 3-runs-per-method noise study. | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "react", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:927665", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-21T23:42:32.510205+00:00 | null | null | null | null | null | null | -1 | 2026-08-21T23:42:32.510205+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-react-239q-v3 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench react repeat 3 of 3 (resubmit of 927664, bad node), vista job 927709. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 172/239 answered, mean HMS 0.0972 over answered / 0.0700 strict-239. Part of 3-runs-per-method noise study. | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "react", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:927709", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-21T23:42:39.552864+00:00 | null | null | null | null | null | null | -1 | 2026-08-21T23:42:39.552864+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-fold-239q-v2 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench fold repeat 2 of 3, vista job 927666. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 192/239 answered, mean HMS 0.1066 over answered / 0.0856 strict-239. Part of 3-runs-per-method noise study. | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "fold", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:927666", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-21T23:42:46.329121+00:00 | null | null | null | null | null | null | -1 | 2026-08-21T23:42:46.329121+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-fold-239q-v3 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench fold repeat 3 of 3, vista job 927667. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 189/239 answered, mean HMS 0.0989 over answered / 0.0782 strict-239. Part of 3-runs-per-method noise study. | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "fold", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:927667", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-21T23:42:53.245851+00:00 | null | null | null | null | null | null | -1 | 2026-08-21T23:42:53.245851+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-239q-v2 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph repeat 2 of 3, output-discipline prompt, vista job 927668. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 151/239 answered, mean HMS 0.0973 over answered / 0.0615 strict-239. Part of 3-runs-per-method noise study. | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:927668", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-21T23:43:00.127442+00:00 | null | null | null | null | null | null | -1 | 2026-08-21T23:43:00.127442+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-239q-v3 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph repeat 3 of 3, output-discipline prompt, vista job 927669. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 153/239 answered, mean HMS 0.0944 over answered / 0.0604 strict-239. Part of 3-runs-per-method noise study. | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:927669", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-21T23:43:07.177425+00:00 | null | null | null | null | null | null | -1 | 2026-08-21T23:43:07.177425+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-fixedops-239q-v1 | eval_discoverybench_qwen3_30b_instruct_8node.sh + rejudge backfill | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "consolidation_interval": 0, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph with SIX graph-op fixes (prompt example fix, junk-observation filter, explanatory op-failure feedback + eligible-id lists, auto-cleanup notices, feedback slimming) AND forced consolidation OFF (SAB_CONSOLIDATION_INTERVAL=0). vista job 928332, repeat v1. 164/239 answered, strict 0.0731, invalid-o... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph", "qwen3-30b", "fixed-ops"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:928332", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-22T17:35:19.398402+00:00 | null | null | null | null | null | null | -1 | 2026-08-22T17:35:19.398402+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-fixedops-239q-v2 | eval_discoverybench_qwen3_30b_instruct_8node.sh + rejudge backfill | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "consolidation_interval": 0, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph with SIX graph-op fixes (prompt example fix, junk-observation filter, explanatory op-failure feedback + eligible-id lists, auto-cleanup notices, feedback slimming) AND forced consolidation OFF (SAB_CONSOLIDATION_INTERVAL=0). vista job 928333, repeat v2. 152/239 answered, strict 0.0605, invalid-o... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph", "qwen3-30b", "fixed-ops"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:928333", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-22T17:35:28.181936+00:00 | null | null | null | null | null | null | -1 | 2026-08-22T17:35:28.181936+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-staterecovery-239q-v1 | eval_discoverybench_qwen3_30b_instruct_8node.sh + rejudge backfill | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "consolidation_interval": 0, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph FULL stack: six graph-op fixes + consolidation OFF + answer-discipline prompt + STATE-RECOVERY prompt (on truncation treat code as never-run, verify state, patch only missing steps). vista job 928943, repeat v1. 139/239 answered, strict 0.0632. Invalid-op rate down from 63% baseline; answer rate... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph", "qwen3-30b", "fixed-ops"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:928943", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-22T19:23:33.793152+00:00 | null | null | null | null | null | null | -1 | 2026-08-22T19:23:33.793152+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-staterecovery-239q-v2 | eval_discoverybench_qwen3_30b_instruct_8node.sh + rejudge backfill | Qwen/Qwen3-30B-A3B-Instruct-2507 | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "consolidation_interval": 0, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph FULL stack: six graph-op fixes + consolidation OFF + answer-discipline prompt + STATE-RECOVERY prompt (on truncation treat code as never-run, verify state, patch only missing steps). vista job 928944, repeat v2. 144/239 answered, strict 0.0758. Invalid-op rate down from 63% baseline; answer rate... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph", "qwen3-30b", "fixed-ops"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:928944", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-22T19:23:41.308400+00:00 | null | null | null | null | null | null | -1 | 2026-08-22T19:23:41.308400+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-8b-239q-v1 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph-8b Qwen3-8B repeat 1 of 3, full fixed config (six fixes + consolidation off + both prompts); DPO data-generation run, vista job 932387. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 172/239 answered, mean HMS 0.1176 over answered / 0.0846 strict-239. Part of DPO data generation: 8B strict ... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph-8b", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932387", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-23T21:22:03.295208+00:00 | null | null | null | null | null | null | -1 | 2026-08-23T21:22:03.295208+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-8b-239q-v2 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph-8b Qwen3-8B repeat 2 of 3, same config, vista job 932388. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 177/239 answered, mean HMS 0.1051 over answered / 0.0778 strict-239. Part of DPO data generation: 8B strict scores 0.0846/0.0778/0.0832 (above all 30B ctxgraph runs 0.060-0.076 and on pa... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph-8b", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932388", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-23T21:22:11.209010+00:00 | null | null | null | null | null | null | -1 | 2026-08-23T21:22:11.209010+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-8b-239q-v3 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph-8b Qwen3-8B repeat 3 of 3, same config, vista job 932389. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 178/239 answered, mean HMS 0.1116 over answered / 0.0832 strict-239. Part of DPO data generation: 8B strict scores 0.0846/0.0778/0.0832 (above all 30B ctxgraph runs 0.060-0.076 and on pa... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph-8b", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932389", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-23T21:22:19.608840+00:00 | null | null | null | null | null | null | -1 | 2026-08-23T21:22:19.608840+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-react-8b-239q-v1 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench react-8b Qwen3-8B react baseline repeat 1/3, original shared config; strict 0.0584, vista job 932456. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 121/239 answered, mean HMS 0.1153 over answered / 0.0584 strict-239. Part of the 8B fair three-way + DPO data generation batch (2026-08-23/24). | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "react-8b", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932456", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T01:59:28.580805+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T01:59:28.580805+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-react-8b-239q-v2 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench react-8b Qwen3-8B react baseline repeat 2/3; strict 0.0555, vista job 932457. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 131/239 answered, mean HMS 0.1013 over answered / 0.0555 strict-239. Part of the 8B fair three-way + DPO data generation batch (2026-08-23/24). | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "react-8b", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932457", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T01:59:37.345349+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T01:59:37.345349+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-react-8b-239q-v3 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench react-8b Qwen3-8B react baseline repeat 3/3; strict 0.0575, vista job 932458. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 132/239 answered, mean HMS 0.1041 over answered / 0.0575 strict-239. Part of the 8B fair three-way + DPO data generation batch (2026-08-23/24). | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "react-8b", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932458", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T01:59:46.998974+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T01:59:46.998974+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-fold-8b-239q-v1 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench fold-8b Qwen3-8B fold baseline repeat 1/3, original shared config; strict 0.0733, vista job 932459. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 155/239 answered, mean HMS 0.1131 over answered / 0.0733 strict-239. Part of the 8B fair three-way + DPO data generation batch (2026-08-23/24). | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "fold-8b", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932459", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T01:59:54.477213+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T01:59:54.477213+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-fold-8b-239q-v2 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench fold-8b Qwen3-8B fold baseline repeat 2/3; strict 0.0813, vista job 932460. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 144/239 answered, mean HMS 0.1349 over answered / 0.0813 strict-239. Part of the 8B fair three-way + DPO data generation batch (2026-08-23/24). | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "fold-8b", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932460", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T02:00:03.081551+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T02:00:03.081551+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-fold-8b-239q-v3 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench fold-8b Qwen3-8B fold baseline repeat 3/3; strict 0.0728, vista job 932461. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 150/239 answered, mean HMS 0.1160 over answered / 0.0728 strict-239. Part of the 8B fair three-way + DPO data generation batch (2026-08-23/24). | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "fold-8b", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932461", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T02:00:10.771908+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T02:00:10.771908+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-8b-clean-239q-v1 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph-8b-clean Qwen3-8B ctxgraph FAIR config (six graph-op fixes + consolidation off; generic prompts removed) repeat 1/3; strict 0.0709, vista job 932506. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 149/239 answered, mean HMS 0.1138 over answered / 0.0709 strict-239. Part of the 8B fair three... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph-8b-clean", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932506", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T02:00:19.449197+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T02:00:19.449197+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-8b-clean-239q-v2 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph-8b-clean Qwen3-8B ctxgraph fair config repeat 2/3; strict 0.0646, vista job 932507. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 157/239 answered, mean HMS 0.0983 over answered / 0.0646 strict-239. Part of the 8B fair three-way + DPO data generation batch (2026-08-23/24). | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph-8b-clean", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932507", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T02:00:27.612682+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T02:00:27.612682+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-8b-clean-239q-v3 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph-8b-clean Qwen3-8B ctxgraph fair config repeat 3/3; strict 0.0681, vista job 932508. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 152/239 answered, mean HMS 0.1071 over answered / 0.0681 strict-239. Part of the 8B fair three-way + DPO data generation batch (2026-08-23/24). | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph-8b-clean", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932508", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T02:00:36.269566+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T02:00:36.269566+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-8b-synth-239q-v1 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph-8b-synth Qwen3-8B ctxgraph on 200 SYNTH tasks (DPO training data source, zero dataset overlap with real eval); strict 0.1346, vista job 932521. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 150/239 answered, mean HMS 0.1794 over answered / 0.1126 strict-239. Part of the 8B fair three-way +... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph-8b-synth", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932521", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T02:00:44.256238+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T02:00:44.256238+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-8b-synth-239q-v2 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph-8b-synth synth repeat 2/3; strict 0.1250, vista job 932522. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 149/239 answered, mean HMS 0.1678 over answered / 0.1046 strict-239. Part of the 8B fair three-way + DPO data generation batch (2026-08-23/24). | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph-8b-synth", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932522", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T02:00:53.496320+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T02:00:53.496320+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-8b-synth-239q-v3 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen/Qwen3-8B | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph-8b-synth synth repeat 3/3; strict 0.1502, vista job 932523. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 152/239 answered, mean HMS 0.1976 over answered / 0.1257 strict-239. Part of the 8B fair three-way + DPO data generation batch (2026-08-23/24). | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph-8b-synth", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:932523", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T02:01:02.453810+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T02:01:02.453810+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-8b-dpo-239q-v1 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen3-8B + DPO LoRA (merged; trained on synth-task pairs, zero overlap with eval set) | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph-8b-dpo Qwen3-8B + DPO (LoRA r16, 147 synth pairs, 3 epochs, beta 0.1) eval 1/2 on 239 REAL tasks; strict 0.0778, answered 156, zero truncation-retry prompts, vista job 933234. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 156/239 answered, mean HMS 0.1193 over answered / 0.0778 strict-239.... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph-8b-dpo", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:933234", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T04:08:05.782356+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T04:08:05.782356+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-8b-dpo-239q-v2 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen3-8B + DPO LoRA (merged; trained on synth-task pairs, zero overlap with eval set) | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph-8b-dpo DPO eval 2/2; strict 0.0762, answered 156, vista job 933235. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 156/239 answered, mean HMS 0.1168 over answered / 0.0762 strict-239. Part of DPO data generation: 8B strict scores 0.0846/0.0778/0.0832 (above all 30B ctxgraph runs 0.060-0.076... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph-8b-dpo", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:933235", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-24T04:08:13.790335+00:00 | null | null | null | null | null | null | -1 | 2026-08-24T04:08:13.790335+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-8b-dpo-v2-239q-v1 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen3-8B + DPO LoRA (merged; trained on synth-task pairs, zero overlap with eval set) | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph-8b-dpo-v2 DPO v2 (87 quality-filtered pairs, beta 0.1 by tie-rule, 2 epochs) eval 1/2; strict 0.0752, answered 157, vista job 936544. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 157/239 answered, mean HMS 0.1144 over answered / 0.0752 strict-239. Part of DPO data generation: 8B strict sc... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph-8b-dpo-v2", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:936544", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-25T01:08:50.065012+00:00 | null | null | null | null | null | null | -1 | 2026-08-25T01:08:50.065012+00:00 |
browsecomp-ctxgraph-30b-rl-discoverybench-ctxgraph-8b-dpo-v2-239q-v2 | eval_discoverybench_qwen3_30b_instruct_8node.sh (+ rejudge backfill for rate-limited judgings) | Qwen3-8B + DPO LoRA (merged; trained on synth-task pairs, zero overlap with eval set) | {"max_turn": 24, "response_length": 8192, "turn_max_new_tokens": 1024, "judge": "gpt-5-nano (azure)"} | ["discoverybench real test split (239 queries)"] | DiscoveryBench ctxgraph-8b-dpo-v2 DPO v2 eval 2/2; strict 0.0758, answered 163, vista job 936545. 239 queries, max_turn=24, judge gpt-5-nano (Azure). 163/239 answered, mean HMS 0.1111 over answered / 0.0758 strict-239. Part of DPO data generation: 8B strict scores 0.0846/0.0778/0.0832 (above all 30B ctxgraph runs 0.060... | ["browsecomp-ctxgraph-30b-rl", "discoverybench", "ctxgraph-8b-dpo-v2", "qwen3-30b", "repeat-run"] | {"experiment_name": "browsecomp-ctxgraph-30b-rl", "job_id": "vista:936545", "cluster": "vista", "artifact_status": "final", "canary": false} | 2026-08-25T01:08:58.918363+00:00 | null | null | null | null | null | null | -1 | 2026-08-25T01:08:58.918363+00:00 |
RACA-PROJECT-MANIFEST
Central registry of all datasets in the lingchensanwen organization.
- Total Datasets Tracked: 41
- Last Updated: 2026-08-25T01:09:01.008403+00:00
Usage
from datasets import load_dataset
manifest = load_dataset("lingchensanwen/RACA-PROJECT-MANIFEST", split="train")
print(f"Tracking {len(manifest)} datasets")
- Downloads last month
- 131