A student scorer from MIRA (Mid-training Rubric Anchoring for Source-Aware Data Selection), fine-tuned to score software-engineering repository-repair agents along a group-specific set of anchor rubric dimensions.

📄 Paper: MIRA: Mid-training Rubric Anchoring for Source-Aware Data Selection (EMNLP 2026) 💻 Code: https://github.com/Multilingual-Multimodal-NLP/mira

TL;DR

MIRA is a source-aware data selection framework for heterogeneous mid-training corpora. Instead of applying a single global quality rubric, MIRA (1) clusters sources into capability-coherent groups, (2) lets a frontier teacher (Kimi-K2.6) freely propose rubric dimensions and anchors them per group, (3) distills the anchored teacher into a lightweight per-group student scorer, and (4) applies reliability-aware aggregation with per-source retention thresholds.

This repository is one of those student scorers — variant 4 in the Agent family, specialized for software-engineering repository-repair agents. Given an in-distribution record, it produces a numerical score and a short rationale for every anchor dimension in this group's rubric.

Model summary


Architecture	Mixture-of-Experts decoder (35B total / ≈3B active params)
Base model	Qwen3.5-35B-A3B-Base
Fine-tuning	Full-parameter SFT on Kimi-K2.6 anchored teacher labels
Domain	Repository-level software-engineering repair agents (OpenHands, daVinc-Dev, swe_qa) — issue → patch trajectories with test-driven verification and minimal-diff style.
Anchor rubric	15 group-specific dimensions (`group_D_dim_anchors.jsonl` in the project repo)
Source count	3 agent sources
Phase-2 corpus (this group)	231,762 teacher-scored records
Output	Structured (score, rationale) per anchor dimension
Precision	BF16
License	Apache-2.0 (inherits from Qwen3)

Sources covered

This scorer is calibrated for the following mid-training sources in the Agent / SWE repository repair group:

Source	Description
`agent_openhands`	OpenHands SWE agent traces
`davinc_dev`	daVinc-Dev env-native pass+fail traces
`swe_qa`	SFT SWE-bench issue + tool-call traces

The full source-grouping report (KMeans k=4 / 5 clusters, intra-group cosine similarities) is in the project repo.

Anchor dimensions (15 slots)

The scoring rubric for this group, discovered via Kimi-K2.6 free-form judging and clustered into 15 anchor dimensions (KMeans k=15 over the group's dim-score embeddings). Names are read verbatim from group_D_dim_anchors.jsonl. Dimensions are sorted by cluster size — larger clusters dominate the corpus and carry more signal. When two clusters' centroids collided on the same anchor name, only the cluster that holds the most records labeled with that name keeps the bare form; the other gets a parenthetical sub-label drawn from its own name distribution so all 15 slots are uniquely identifiable.

Slot	Dimension	Cluster size
A1	Plan Quality / Reasoning Transparency (Goal Achievement)	59,467
A2	Output Style/Communication	41,355
A3	Goal Achievement	38,129
A4	Plan Quality / Reasoning Transparency	36,330
A5	Test Coverage & Verification Rigor	29,347
A6	Multi-turn Coherence	29,127
A7	Error Recovery	28,843
A8	Tool Argument Correctness	28,122
A9	Action Efficiency	27,440
A10	User Intent Comprehension	24,691
A11	Safety & scope	24,339
A12	Observation handling	23,739
A13	Code Change Minimality	20,203
A14	Tool Selection	17,000
A15	Tool Selection (Safety & Scope)	16,168

The scorer outputs one [Ai] <dimension>: <score>/10 — <rationale> line per slot, plus overall, training_recommendation, domain_tag, and brief.

Where this model fits in the MIRA pipeline

┌──────────────────┐  ┌──────────────────┐  ┌──────────────────┐  ┌──────────────────┐
│ 1. Rubric        │  │ 2. Anchored      │  │ 3. Reliability   │  │ 4. Data          │
│    Discovery     │→ │    Judge         │→ │    Aggregation   │→ │    Selection     │
│ (Kimi-K2.6,      │  │    Distillation  │  │ (mask unreliable │  │ (per-source      │
│  free-form       │  │ ◀── THIS MODEL   │  │  src×dim cells)  │  │  retention)      │
│  judging)        │  │                  │  │                  │  │                  │
└──────────────────┘  └──────────────────┘  └──────────────────┘  └──────────────────┘

MIRA-Agent-Group4 lives in Stage 2: it scores the full Agent / SWE repository repair corpus so that downstream stages can apply reliability masking and source-aware retention.

Intended use

Primary: Score software-engineering repository-repair agents on this group's anchor dimensions to drive source-aware data selection and filtering.
Secondary: Research on rubric distillation, semantic quality scoring, and reliability diagnostics for heterogeneous training corpora.

Not intended for:

General-purpose chat or instruction following — fine-tuned to emit structured scores, not freeform dialogue.
Single-shot quality judgments without the anchor-dimension prompt template — outputs will be miscalibrated.
Records outside the Agent / SWE repository repair group; use the matching sibling scorer instead.

Deployment

The scorer is designed to be served via vLLM behind an OpenAI-compatible endpoint and called in batch from the MIRA scoring pipeline.

1. Serve with vLLM (recommended)

bash

vllm serve whw06/MIRA-Agent-Group4 \
    --tensor-parallel-size 8 \
    --dtype bfloat16 \
    --max-model-len 65536 \
    --max-num-batched-tokens 131072 \
    --gpu-memory-utilization 0.9 \
    --trust-remote-code \
    --port 8000

Why these values (verified on H200 141GB during the paper's per-source evaluation):

max-model-len=65536 — 2× the mid-training cutoff. Records can hit ~60K tokens for densely-tokenized sources; 40K runs into prompt-overflow errors.
max-num-batched-tokens=131072 — supports two full-length sequences per scheduling step.
gpu-memory-utilization=0.9 — 35B BF16 weights take ~70GB, leaving ~57GB KV cache. Roughly 4 concurrent 65K-context sequences per GPU.
8-way tensor parallel works well for the 35B MoE on a single 8×H200/A100 node.

2. Call from Python

python

from openai import OpenAI

client = OpenAI(base_url="http://localhost:8000/v1", api_key="EMPTY")

resp = client.chat.completions.create(
    model="whw06/MIRA-Agent-Group4",
    messages=[
        {"role": "system", "content": SYSTEM_PROMPT},   # group-D anchor calibration
        {"role": "user",   "content": USER_PROMPT},     # record + [A1]..[A15] template
    ],
    temperature=0.7,
    top_p=0.95,
    max_tokens=2048,
)
print(resp.choices[0].message.content)

3. Prompt template

The user message asks for one structured line per anchor dimension (top-15 of this group):

[A1] {anchor_dim_1}: <score>/10 — <justification>
[A2] {anchor_dim_2}: <score>/10 — <justification>
...
[A15] {anchor_dim_15}: <score>/10 — <justification>
overall: <0-100>
training_recommendation: <keep | downsample | drop>
domain_tag: <short tag>
brief: <one-sentence summary>

The system prompt embeds the top-12 anchor calibration references (canonical examples from clustering) so the student matches the teacher's scoring scale. The full prompt builder, anchor JSONL files, and output parser are in the project repo's scoring/score_agent_anchored.py.

Training details


Teacher	Kimi-K2.6 (free-form rubric discovery in Phase 1; anchored re-scoring in Phase 2)
Training data	Kimi-K2.6 anchored labels on this group's Phase-2 corpus, split into a distillation set + a held-out validation split for reliability diagnostics
Loss	Standard next-token CE over (score, rationale) labels for every anchor dimension
Hyperparameters	Held constant across all MIRA student scorers; full settings in paper Appendix A.4
Validation	Per-dimension teacher–student MAE and Spearman ρ on a held-out split; dimensions failing reliability thresholds are masked post-hoc (Figure 3 in the paper)

Training loss / step curve is preserved in trainer_state.json for full reproducibility.

Headline results (from the paper)

End-to-end downstream evaluation: Qwen2.5-Coder-14B mid-trained on 25B-token MIRA-selected subsets vs. baselines, then SFT, evaluated on 9 code benchmarks across 4 categories.

Method	Code Gen	MultiplE	SQL (EX)	SWE-Multi	Macro Avg
Base + SFT (no mid)	53.91	72.57	64.24	3.67	48.60
Raw Mixture (50B)	53.71	67.42	94.18	40.00	63.83
Random (25B)	52.71	71.44	91.03	35.00	63.23
DataMan (25B)	53.82	71.38	93.84	33.00	63.01
DSIR (25B)	48.74	67.26	95.20	27.00	59.55
PPL (25B)	50.52	57.74	90.66	20.00	54.73
MIRA-Global (25B)	53.12	67.84	94.26	32.00	61.81
MIRA-Group (25B)	54.53	71.85	94.08	36.33	64.20
MIRA-Source (25B)	54.18	72.84	94.38	30.33	62.93

MIRA-Group matches the full 50B-token raw mixture while using only half the tokens, and out-performs all 25B-token selection baselines on the macro average. This scorer is one of the 12 student models used by the MIRA-Group variant.

Sibling models

MIRA releases one student scorer per source-group variant. Use the matching scorer for each record's format:

Agent: whw06/MIRA-Agent-Group1 · -Group2 · -Group3 · MIRA-Agent-Group4 (this model)
QA: whw06/MIRA-QA-Group1 · -Group2 · -Group3 · -Group4 · -Group5
Text: whw06/MIRA-Text-Group1 · -Group2 · -Group3

Limitations

MIRA addresses source-aware filtering only. Source discovery, mixture-ratio design, curriculum scheduling, deduplication and contamination control remain orthogonal concerns.
This scorer is calibrated against the Agent / SWE repository repair group; cross-domain transfer is not advised — use the matching sibling for other source formats.
Some anchor dimensions exhibit high teacher–student MAE and are masked post-hoc during aggregation (see paper §3.4). The model still emits scores for masked dimensions; downstream consumers should re-apply the reliability mask from the project repository.
Calibrated on 3 sources within this group; behavior on out-of-distribution formats is unverified.

Citation

bibtex

@inproceedings{wang2026mira,
  title     = {MIRA: Mid-training Rubric Anchoring for Source-Aware Data Selection},
  author    = {Wang, Haowen and Du, Yaxin and Yang, Jian and Wu, Jiajun and
               Liu, Shukai and Zhang, Yuxuan and Wang, Pingjie and Chen, Siheng and
               Zheng, Tuney and Zhou, Ming and Liu, Xianglong},
  booktitle = {Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing (EMNLP)},
  year      = {2026}
}

Acknowledgments

Built on Qwen3.5-35B-A3B-Base and the Megatron-LM training stack. Teacher labels generated with Kimi-K2.6.